c0mpiled-15: Penn State
Overview
- 347 registrations, 68 projects submitted by 180+ builders, and 17 prizes awarded.
- What stood out most was the PSU community. Alumni founders, some who graduated a couple of years ago and others two decades ago, came back to teach, mentor, and cheer on students, and their energy carried through the whole event.
- None of this would have happened without the support of Penn State’s Engineering Entrepreneurship Department and The Builder’s Collective, who helped make the hackathon run as well as it did!
Reflections
Good
- Participants used the sponsor stack heavily. Of our 68 participants, 35 used the Luminal cloud, followed closely by MongoDB (28), AWS (27) and AGI Inc (27). Many used more than one sponsor tool, so these numbers overlap.
- The Luminal challenge pushed teams past demos and into real measurement, with many submissions reporting held-out results.
- We allocated 8 full hours of build time, which converted into 68 submissions, the largest amount we have gotten!
- Lots of new and solo builders: 15 projects came from first-time hackers and 19 were built solo.
Bad
- Judging took longer than planned due to the number of submissions our track received. Good problem to have!
Sponsors & Judges
68 submissions, 17 prizes awarded! Hosted by c0mpiled and Builder's Collective.
Sponsors: Transpose Platform, Luminal (YC S25), AWS, MongoDB, Bloomerang, AGI Inc, Engineering Entrepreneurship Program, Red Bull
Judges
- Joe Fioti — Luminal (YC S25), PSU alum
- Amanda Lind — Unseen AI (SPC), PSU alum
- Kevin Cohen — VetRec (YC S23), PSU alum
- Het Sheth — Labs @ Bloomerang, PSU alum
- Professor Ted Graef — Director, Engineering Entrepreneurship
- Professor Greg Woodman — Adjunct Professor, Engineering Entrepreneurship
- Kanika Gupta — Founding Engineer, Raya (HF0 W26)
- Kartikey Pandey — MTS, Cara
- Krushil Amrutiya — SWE, Fanatics
- Ajay Narayanan Sridhar — ML Engineer, Dropbox
- Vivek Patel — SWE, AGI
- Rushi Patel — Co-Founder, Starborn
Prizes & Recognition
Transpose Platform — Self-Improving Agent Teams (Luminal) & YC Requests for Startups
1st Place: Scam Shield ↓ entry — Scam Shield is on-device AI that listens to a phone call on speaker and warns older adults in real time when it sounds like a scam, explaining why in plain English and suggesting how to check who’s really calling. Speech-to-text and a local Llama model run entirely on the laptop, so it works with no internet and no audio ever leaves the device.
Anushka Shrivastava
GitHub: https://github.com/itsanushkah/scam-shield | Video: https://drive.google.com/drive/folders/1lCx-A_Ygb6mJR8M349jjeJGy7eFjMhUb?usp=sharing
2nd Place: Digital Twin Security ↓ entry — Digital Twin Security tests the software that drives autonomous machines inside a physics simulation, so teams find how their system fails before it's placed in production with people. Every failure comes with the evidence, the likely cause and a replay you can watch in 3D.
SJ Bae, Alex Safai, Julian Safai
GitHub: https://github.com/AlexSafai/digital-twin-security | Video: https://www.youtube.com/watch?v=eTJcPYpgx2s
3rd Place: Le-Quant Labs ↓ entry — We built an autonomous, self-improving quantitative research firm powered by nine specialized AI agents that collaborate to discover market anomalies, test hypotheses, validate strategies, and evaluate trading opportunities. Unlike traditional AI trading systems, our agents analyze their own performance, identify weaknesses, and improve their research processes—demonstrating measurable gains in stock-market crash prediction accuracy through controlled experimentation.
Ryan Bayon, David Sepko, Thomas Askharoun
GitHub: https://github.com/askharoun/Le-quant | Video: https://youtu.be/sDySKLZVmmw
AWS — Best Use of AWS
1st Place: OrgFlow ↓ entry — OrgFlow is an AI-powered knowledge management platform that helps nonprofits preserve institutional knowledge, connect documents from services like Google Drive, and turn scattered information into searchable answers, actionable procedures, and visual knowledge maps. With voice assistance, source-backed AI responses, and a volunteer coordinator agent, OrgFlow helps teams work efficiently and ensures that when someone leaves, what they know stays. (Also 2nd in the Bloomerang track.)
Aiden Johnson, Ryan Nair, Parth Chogule, Samarth Singh
GitHub: https://github.com/tgmstudios/OrgFlow | Video: https://www.youtube.com/watch?v=enjrHv29DNI | Live: https://orgflow.dev
2nd Place: Roster ↓ entry — Students can't get hired, clubs can't get funding, companies can't find talent. Roster is one service that solves all three problems.
Adam Torres Encarnacion
GitHub: https://github.com/0xABAN/c0mpiled | Video: https://youtu.be/g0Ngz6rD-lY?si=S3DNa9XmIlRAcktf
3rd Place: Levee ↓ entry — Levee helps first-time buyers figure out how much they really need to own a rental property, then stress-tests a real home against a bad flood year using FEMA flood data. Every number comes from a deterministic calculator, and Amazon Bedrock only explains the results in plain English.
Vitaliy Pikalo, Alex Hasenbien
GitHub: https://github.com/AHasenbein/Levee | Video: https://youtu.be/zIVy-iNLBfk?is=LubGQjQhQwZG7ehC
Engineering Entrepreneurship — Coolest Sales Story
1st Place: InstaQuote (InstaQuotes) ↓ entry — InstaQuote gives landscaping and lawn-care companies an embeddable widget that turns a customer’s address into an instant, accurate quote. It measures the yard from satellite imagery and prices the job from the contractor’s own rates. Customers get a price in seconds without waiting for a site visit, and businesses save the time and fuel of driving out to measure every job.
Gus Platt, Austin Strickler, Amanda Dettmar, Jackson McCarten, Keevan Nguyen
GitHub: https://github.com/jacksonmccarten/lawn-estimate.git | Video: https://www.youtube.com/watch?v=1WCOEH5n5uU
2nd Place: CallCrowd (Pairadox) ↓ entry — CallCrowd helps independent local businesses like restaurants, cafés, salons and shops, which can’t afford consultants and end up deciding prices, hours and promos on gut feel. An owner types their business name and a question, and our agents turn the business’s real reviews, menu and competitors into 50 AI customers who debate the decision with cited evidence. A judge agent then gives a clear verdict, a plan and a print-ready flyer for $19, built on AWS and MongoDB so it gets smarter about each business with every question.
Atharv Mittal, Biyash Das
GitHub: https://github.com/atharv109/compiled15 | Video: https://www.loom.com/share/7e4e6045f64a41879b1164079384a505
3rd Place: Vitals ↓ entry — Vitals scores every incoming lead for service businesses by its likely job value and chance of closing, and compares the expected work against the team's capacity to show whether the schedule will fill. It tells owners what to do next — get more leads, hold steady, or raise prices — before the schedule empties.
David Ulmer, Shanin Drucquer, Alejandro Vidal, Sergei Zhdanov, Saril Pattnaik
GitHub: https://github.com/Flizzy50/Business-Pulse-Compiled | Video: https://www.loom.com/share/4b4b29ffddd0447ea8022a404de2c9d2 | Live: https://business-pulse-compiled.vercel.app
Bloomerang — An AI Agent for Volunteer Coordinators
1st Place: CallCoach ↓ entry — CallCoach is a calm workbench that helps volunteer calling teams plan who to call, prepare each caller, record what happened, and follow through. A person approves every step, and the team's history stays with the team when people rotate off.
Ved Patel, Mat Muthambi
GitHub: https://github.com/matmanna/callcoach | Video: https://youtu.be/NcYWvT19fdA
2nd Place: OrgFlow ↓ entry — OrgFlow is an AI-powered knowledge management platform that helps nonprofits preserve institutional knowledge, connect documents from services like Google Drive, and turn scattered information into searchable answers, actionable procedures, and visual knowledge maps. With voice assistance, source-backed AI responses, and a volunteer coordinator agent, OrgFlow helps teams work efficiently and ensures that when someone leaves, what they know stays. (Also 1st in the AWS track — see above.)
Aiden Johnson, Ryan Nair, Parth Chogule, Samarth Singh
GitHub: https://github.com/tgmstudios/OrgFlow | Video: https://www.youtube.com/watch?v=enjrHv29DNI | Live: https://orgflow.dev
3rd Place: Tari ↓ entry — Tari is an iMessage-native secretary for nonprofit volunteer coordinators, backed by an agent loop that handles forms and drafts outbound messages for the coordinator to approve. (Also the MICS Solo Hacker winner.)
Emilia Palmer
GitHub: https://github.com/emipalmer/coordinator-assistant | Video: https://www.youtube.com/watch?v=LjrNd8nAadE
AGI Inc — Best Use of On-Device Intelligence
1st Place: Gulag Loop ↓ entry — Gulag Loop is an AI system that learns to play Call of Duty Zombies using a vision model, MongoDB memory, and four Luminal-powered strategy agents. A live dashboard shows gameplay, agent recommendations, and results as the system tests new strategies to improve performance — data that could be used to train robots and other systems for urban combat situations.
Rishit Rai, Deekshith Reddy, Jeffrey Gao
GitHub: https://github.com/Dee-1862/Gulag-Loop | Video: https://cap.so/s/8set1rw3q9z79rv | Live: https://zombieloop-dashboard.onrender.com/
2nd Place: Scam Shield ↓ entry — Scam Shield is on-device AI that listens to a phone call on speaker and warns older adults in real time when it sounds like a scam, explaining why in plain English and suggesting how to check who’s really calling. Speech-to-text and a local Llama model run entirely on the laptop, so it works with no internet and no audio ever leaves the device. (Also 1st in the Transpose Platform track — see above.)
Anushka Shrivastava
GitHub: https://github.com/itsanushkah/scam-shield | Video: https://drive.google.com/drive/folders/1lCx-A_Ygb6mJR8M349jjeJGy7eFjMhUb?usp=sharing
MongoDB — Best Use of MongoDB
1st Place: NextUp ↓ entry — NextUp is an AI agent for volunteer coordinators. It handles cancellations, coverage gaps, plain-English requests and weather risks: it finds available, fairly-rested volunteers or alternative plans, and drafts the messages. Nothing reaches a volunteer until the coordinator approves, so the agent does the work while the coordinator stays in charge.
Woojoo Song, Woosung Lim
GitHub: https://github.com/CosmosBoy112/next_up | Video: https://youtu.be/J5t3U6xgwqs
Winner: TokenForge ↓ entry — TokenForge is a self-improving agent team: six DeepSeek-V4.1-Flash agents on Luminal rewrite the core matrix-multiply kernel of an on-device LLM (llama2.c on an M3 Pro), while a harness they can’t modify keeps only kernels that produce byte-identical output, and MongoDB Atlas vector search feeds every past win and failure into the next round. On held-out models and prompts never used during optimization, the final kernel generates text 7.82× faster than stock llama2.c (13.9 → 110 tok/s on the 110M model).
Aryaman Singh Dev
GitHub: https://github.com/dev4-gpt/tokenforge | Video: https://youtu.be/2ZHimN2vUzo
MICS — For Solo Hackers!
Winner: Tari ↓ entry — Tari is an iMessage-native secretary for nonprofit volunteer coordinators, backed by an agent loop that handles forms and drafts outbound messages for the coordinator to approve. (Also 3rd in the Bloomerang track — see above.)
Emilia Palmer
GitHub: https://github.com/emipalmer/coordinator-assistant | Video: https://www.youtube.com/watch?v=LjrNd8nAadE
get the next recap, and the date of the next hackathon.
one email per event. winners, every submission, what we learned. no spam.
All Submissions
AirDrums (Transpose (YC RFS) · Transpose (Luminal) · MongoDB)#
Description: 3d visuals along with an AI tutor and progressive game that allows anybody to play the drums with only a camera, open space, and their hands.
- Team: Shaurya Pamarthi, Veer Gandhi, Stefan Ghiugan, Aayush Lamichhane
- GitHub: https://github.com/shauryapamarthi-netizen/hackathon-backend
- Video: https://www.youtube.com/watch?v=AR9B3-dKQiU
alrminator (Transpose (Luminal) · AGI Inc · MongoDB · MICS)#
Description: A team of parallel AI agents that rewrites its own routine predictor round by round, then runs it on-device to plan your day and control your home.
- Team: Sahaj Nanda
- GitHub: https://github.com/crqzyexprees/routine-agent
- Video: https://drive.google.com/file/d/1UlP0ZTMPyJX9wDZSfp05dUtBtX7LtTxA/view?usp=sharing
Are You Human? (Awesome) (AGI Inc)#
Description: Are You Human? stops deepfake and replay scams with random, real-time liveness challenges, verified on-device in your browser so your video never leaves your computer.
- Team: Ryan, Tanush
- GitHub: https://github.com/RJC862/are-you-human
- Video: https://youtube.com/shorts/3r7IVzHaD6A?feature=share
Brainrot Chess (BNN) (MongoDB)#
Description: AI Agents that play chess against one another in a tournament style format and score system that are evaluated against different bots.
- Team: Brij Kalakheti, Nisha Neupane
- GitHub: https://github.com/nishaneupane/brainrot-chess
- Video: https://youtu.be/Ss1zAhScHoA
CallCoach (Transpose (YC RFS) · AWS · AGI Inc · MongoDB · Bloomerang) — Bloomerang 1st#
Description: CallCoach is a calm workbench that helps volunteer calling teams plan who to call, prepare each caller, record what happened, and follow through. A person approves every step, and the team's history stays with the team when people rotate off.
- Team: Ved Patel, Mat Muthambi
- GitHub: https://github.com/matmanna/callcoach
- Video: https://youtu.be/NcYWvT19fdA
CallCrowd (Pairadox) (AWS · AGI Inc · MongoDB · EED · Bloomerang) — Engineering Entrepreneurship 2nd#
Description: CallCrowd helps independent local businesses like restaurants, cafés, salons and shops, which can’t afford consultants and end up deciding prices, hours and promos on gut feel. An owner types their business name and a question, and our agents turn the business’s real reviews, menu and competitors into 50 AI customers who debate the decision with cited evidence. A judge agent then gives a clear verdict, a plan and a print-ready flyer for $19, built on AWS and MongoDB so it gets smarter about each business with every question.
- Team: Atharv Mittal, Biyash Das
- GitHub: https://github.com/atharv109/compiled15
- Video: https://www.loom.com/share/7e4e6045f64a41879b1164079384a505
Carter Squad (AWS · Bloomerang)#
Description: Ai Schedule manager for non profit, auto calls volunteers and figures out backups
- Team: Carter De Streel, Parag Sahani, Jiawei Xi, Tarun N Shetru
- GitHub: https://github.com/KaneXi/c0mpiled-15
- Video: https://drive.google.com/drive/folders/1eY_kS3u10xoh_Yg9a6WnTR5IKBV4SiY9?usp=sharing
code-anime (Transpose (YC RFS))#
Description: code-anime helps solo storytellers turn a reference image and scene description into short anime sequences. Our prototype combines Luminal-generated visual assets, open-weight video generation, and tools for reviewing and assembling shots, demonstrated through a ten-second bedroom scene.
- Team: Nitish Reddy M
- GitHub: https://github.com/reddynitish/code-anime
- Video: https://youtu.be/QreVPrYYzyM
CogniDrive (AWS · MICS · First-Time Hacker)#
Description: This project is a collaborative chat space using Amazon AWS tools to be able to chat collaborate and share an AI in a company with your audience. This is also an attempt at fulfilling the YC RFS call for a "multiplayer moment" in AI.
- Team: Damien
- GitHub: https://github.com/Hyperstop17/c0mpiled-15Submission
- Video: https://youtu.be/wsTkV6JJpS0
CommonThread (Mismatched) (Transpose (YC RFS) · AGI Inc · Bloomerang)#
Description: CommonThread is a shared room where you, Claude and Codex work together in one conversation. The agents hand work to each other explicitly, review exact versions of each other's output, and you can jump in, answer their questions, pause or stop at any point. It turns two separate AI coding tools into a small team you can actually direct and keep track of.
- Team: Krishna Pagrut, Keshav Khandelwal, Arav Jain, Dev Singh, Rushil Kakkad
- GitHub: https://github.com/DevSingh05/CommonThread.git
CrewSync AI (AGI Inc · Bloomerang · MICS)#
Description: CrewSync AI is an AI-powered event and team coordination platform that simplifies event planning, volunteer scheduling, task management, and communication using natural language. Powered by Luminal AI, it helps student organizations, nonprofits, and event coordinators automate repetitive work, identify scheduling conflicts, and manage events efficiently while keeping humans in control.
- Team: Palash Doshi
- GitHub: https://github.com/palashDoshi1/CrewSync
Digital Twin Security (Transpose (YC RFS)) — Transpose Platform 2nd#
Description: Digital Twin Security tests the software that drives autonomous machines inside a physics simulation, so teams find how their system fails before it's placed in production with people. Every failure comes with the evidence, the likely cause and a replay you can watch in 3D.
- Team: SJ Bae, Alex Safai, Julian Safai
- GitHub: https://github.com/AlexSafai/digital-twin-security
- Video: https://www.youtube.com/watch?v=eTJcPYpgx2s
DinoDocket (Transpose (Luminal) · MongoDB · First-Time Hacker)#
Description: DinoDocket is an AI weekly planner that checks whether your work actually fits around your locked commitments, plans how to do it step by step, and only changes your calendar when you approve. Behind it, a team of four LLM agents keeps improving the scheduling strategy, raising its score on 15 unseen student weeks from 69.9 to 77.1 (+10.3%), with every run stored in MongoDB Atlas.
- Team: Younsoo Park, Molly Barash, Abrar Islam
- GitHub: https://github.com/danielsoo/dinodocket
- Video: https://youtu.be/Txd1QufQKgE?si=dNh44_2l4xNTeZ_m
DispatchIQ (Transpose (Luminal))#
Description: DispatchIQ is an AI-powered field service optimization platform that dynamically reschedules technicians, prioritizes emergency jobs, and minimizes operational disruptions in real time. Powered by self-improving AI agents and constraint-based optimization, it continuously evolves scheduling strategies to reduce delays, travel costs, and downtime.
- Team: Het Patel
- GitHub: https://github.com/dev4-gpt/DispatchIQ.git
- Video: https://www.loom.com/share/5e733a54748e4f82809014e4f38a7b93
FieldDelta (Transpose (YC RFS) · AGI Inc · MICS)#
Description: FieldDelta is a persistent memory layer for physical spaces: scan a site from an iPhone, and it remembers objects across inspections, detects what was added, removed, moved, or unchanged, and preserves operator corrections as site memory. It combines local computer vision with Luminal-powered cloud validation to turn ordinary camera scans into a persistent history of real-world change.
- Team: Yiyao Ding
- GitHub: https://github.com/neod6/FieldDelta
- Video: https://youtu.be/_yurOvLJY5I
Firefly (AGI Inc · MongoDB)#
Description: Firefly is an operating system for mixed physical workforces: AI agents, autonomous yard trucks and people work from one shared record to recover from exceptions at a distribution-center yard, with every step validated and every outcome verified. That record, stored in MongoDB, lets any shift be replayed and lets the agents use recorded outcomes from earlier shifts in their next decision.
- Team: Muhammad Usman, Rayyan Syed
- GitHub: https://github.com/usmannsiddiqui/firefly
- Video: https://youtu.be/GN5DMxZcOPA
GazeShift (Anoop and Hussain) (Transpose (YC RFS) · First-Time Hacker)#
Description: GazeShift is an experimental assistive-vision prototype for older adults with impaired vision. You draw around part of a live camera view and move a clear copy of it wherever you can see best. It’s built with Swift, SwiftUI, ARKit and custom Metal shaders, and it runs entirely on the iPhone. We’re working toward AR/VR wearables with eye tracking and a personal vision map, especially for central vision loss.
- Team: Anoop I, Hussain
- GitHub: https://github.com/Hussain-coder-eng/gazeshift
- Video: https://youtu.be/sY6ck3Sh1E8
GPSI (Transpose (YC RFS) · Transpose (Luminal) · AWS · EED · MICS)#
Description: General Purpose Scientific Intelligence: an AI discovery engine.
- Team: Amartya Sai Andra
- GitHub: https://github.com/cars3/General-Purpose-Scientific-Intelligence
- Video: https://youtu.be/V0-yyAT5Qew
Gulag Loop (Transpose (Luminal) · AGI Inc · MongoDB) — AGI Inc 1st#
Description: Gulag Loop is an AI system that learns to play Call of Duty Zombies using a vision model, MongoDB memory, and four Luminal-powered strategy agents. A live dashboard shows gameplay, agent recommendations, and results as the system tests new strategies to improve performance — data that could be used to train robots and other systems for urban combat situations.
- Team: Rishit Rai, Deekshith Reddy, Jeffrey Gao
- GitHub: https://github.com/Dee-1862/Gulag-Loop
- Video: https://cap.so/s/8set1rw3q9z79rv
Handoff (Floor2) (Transpose (YC RFS) · AWS · AGI Inc)#
Description: Handoff is an on device AI assistant that turns maintenance technicians’ spoken notes into evidence linked reports, preserving corrections and tracking unfinished work. It runs locally without internet, helping the next technician understand what was done, what’s still wrong, and what needs attention.
- Team: Sreenidh Bonagiri, Aditya Turaga
- GitHub: https://github.com/sreenidhbonagiri/handoff
- Video: https://www.youtube.com/watch?v=r0bODlr8o7g
Headway (Transpose (YC RFS) · Transpose (Luminal) · AGI Inc · Bloomerang)#
Description: a toy-diorama transit game whose dispatcher is evolved by AI designers
- Team: Pranav Karra
- GitHub: https://github.com/Pranav-Karra-3301/headway
- Video: https://youtu.be/1A8NY5UbujE
Helix Evolve (Transpose (Luminal) · AGI Inc · MongoDB · First-Time Hacker)#
Description: Helix is a self-improving multi-agent system that evolves code with LLMs through two parallel loops: a wide loop of undirected mutations for diversity, and a hypothesis-driven deep loop that researches why the best programs win and tests each idea with one controlled change. It runs on Luminal Cloud (DeepSeek Flash, with an occasional Pro model for breakthroughs) or fully on-device with a local LLM (Ollama / LM Studio) to keep improving indefinitely at no API cost. On online 2D vector bin packing it cut wasted servers by 26% over First Fit in 30 generations, and kept a 24% cut on held-out workloads it never trained on.
- Team: Shreyas Naik, Saivikram Kamireddy Ravichandran
- GitHub: https://github.com/cybercryptixcoder/helix_evolve
- Video: https://youtu.be/a40rj63gZ_s
Hive (Transpose (YC RFS) · Transpose (Luminal) · AGI Inc · MongoDB)#
Description: Hive is "Google Docs for coding agents": a team shares one live task graph where everyone can watch each AI agent work in real time, redirect it mid-run, take over a task and hand it back with full context. Anyone can catch up on what happened, and approved work is tested and pushed to GitHub with commits from both agents and people.
- Team: Ahnaf Iftikhar, Owen Bress, Jeremy Hsu
- GitHub: https://github.com/OBress/Hive
- Video: https://youtu.be/ntiecI9HVlA
Hormes (Transpose (YC RFS) · AGI Inc)#
Description: Hormes puts an AI companion in your Mac’s notch, helping you reach your academic and professional goals with tools in tutoring and action. It connects your calendar, to-dos, and goals, carrying context across your day to help you tackle what’s next and pick up unfinished work.
- Team: Tazeem Mahashin, Russell Allyn
- GitHub: https://github.com/Taz33m/hormes
- Video: https://youtu.be/ThavmUjLAwU?si=ckDpsDHm6GkXcNwQ
Huddle (Nittany Builders - BC) (Transpose (YC RFS) · Transpose (Luminal) · MongoDB · Bloomerang)#
Description: AI is single-player: one person, one chat box. Huddle is a live shared board where a whole committee and its AI agents work the same queue — agents wake on MongoDB change streams, triage incoming requests on their own, and hand off to whoever claims them. Built for THON, Penn State’s 46-hour dance marathon, where two thousand volunteers currently coordinate through a group chat.
- Team: Tejasv Agarwal, Shivaan Bhalla, Sayan Gupta
- GitHub: https://github.com/sayang7/huddle
- Video: https://drive.google.com/file/d/1Mad2aAYqdm43vQKRzelh-G-V0pNdZWo5/view?usp=sharing
hungry (Transpose (YC RFS) · Transpose (Luminal) · MongoDB · EED · MICS · First-Time Hacker)#
Description: Helps restaurants track the real cost of each dish and create deals without losing money. Students can find and claim affordable food from local restaurants.
- Team: Matthew Thomas
- GitHub: https://github.com/mattothomas/resturant-compiled
- Video: https://drive.google.com/drive/folders/1xlWSVEL5ygjqJTvKCo73-CJZNM77LzZz?usp=drive_link
InstaQuote (InstaQuotes) (AWS · EED) — Engineering Entrepreneurship 1st#
Description: InstaQuote gives landscaping and lawn-care companies an embeddable widget that turns a customer’s address into an instant, accurate quote. It measures the yard from satellite imagery and prices the job from the contractor’s own rates. Customers get a price in seconds without waiting for a site visit, and businesses save the time and fuel of driving out to measure every job.
- Team: Gus Platt, Austin Strickler, Amanda Dettmar, Jackson McCarten, Keevan Nguyen
- GitHub: https://github.com/jacksonmccarten/lawn-estimate.git
- Video: https://www.youtube.com/watch?v=1WCOEH5n5uU
knck2 (Transpose (YC RFS) · Transpose (Luminal) · AGI Inc · MongoDB · EED · MICS)#
Description: on device agent and classification. to make accelerometer data more clean and reliable for real world and ai agent use.
- Team: Devin Myers
- GitHub: https://github.com/Devin-M5706/knck2
- Video: https://youtu.be/vRRy4JN0ol0?is=eLxdYIclQIwwJKn3
Last Call (AWS · EED · First-Time Hacker)#
Description: Last Call is a mobile app and store portal that lets State College restaurants, cafes and bakeries sell what they have left before closing at a discount: students reserve in two taps, get a pickup code, and pay at the counter. Stores list with a single message that AI turns into a ready-to-approve deal, and an AI agent gets anything still unsold to food-rescue volunteers.
- Team: Lakshay Kalra, Chukwukbikem David Dara, Kirubel Tamrat, Taemin Kang
- GitHub: https://github.com/ChukwubikemDara/HelpingYouSellFoodCompile26
- Video: https://drive.google.com/file/d/1Cgv3q5Zbau7jX-cpmebmUaqn8YuLi8tm/view?usp=sharing
Le-Quant Labs (Transpose (YC RFS) · Transpose (Luminal)) — Transpose Platform 3rd#
Description: We built an autonomous, self-improving quantitative research firm powered by nine specialized AI agents that collaborate to discover market anomalies, test hypotheses, validate strategies, and evaluate trading opportunities. Unlike traditional AI trading systems, our agents analyze their own performance, identify weaknesses, and improve their research processes—demonstrating measurable gains in stock-market crash prediction accuracy through controlled experimentation.
- Team: Ryan Bayon, David Sepko, Thomas Askharoun
- GitHub: https://github.com/askharoun/Le-quant
- Video: https://youtu.be/sDySKLZVmmw
Levee (AWS) — AWS 3rd#
Description: Levee helps first-time buyers figure out how much they really need to own a rental property, then stress-tests a real home against a bad flood year using FEMA flood data. Every number comes from a deterministic calculator, and Amazon Bedrock only explains the results in plain English.
- Team: Vitaliy Pikalo, Alex Hasenbien
- GitHub: https://github.com/AHasenbein/Levee
- Video: https://youtu.be/zIVy-iNLBfk?is=LubGQjQhQwZG7ehC
Materially (Transpose (YC RFS) · Transpose (Luminal) · AGI Inc · MongoDB · MICS · First-Time Hacker)#
Description: Materially is a construction takeoff tool built for the AI era. It uses Firecrawl for web search and the JEV decision model for database analysis.
- Team: Ethan Tluszcz
- GitHub: https://github.com/Ethantlz/materially
- Video: https://www.loom.com/share/27b0a2f068ac4fca9b4d22305ab235f5
Munshi (Transpose (Luminal) · AWS · MongoDB · EED)#
Description: Munshi is a self-learning AI back office for local businesses: snap a bill or ring up a sale, and inventory, books and profit update instantly, while an AI manager tells you in plain dollars what to do next. A team of AI agents grades its own predictions against real results and gets smarter with every bill, cutting forecast error 59% on days it had never seen.
- Team: Anant Patel, Meet Bhojani, Vaibhav Mittal, Adwaya Sharma
- GitHub: https://github.com/meetbhojani8989/Munshi
- Video: https://youtu.be/NqYCtDNMmM8
NaviNote (AWS)#
Description: NaviNote is an AI-powered accessibility solution being developed into a lightweight smart cane to help blind and visually impaired individuals navigate their surroundings and manage everyday financial transactions independently. Using computer vision and a conversational AI assistant, it aims to recognize US dollars and UAE dirhams, read receipts, explain purchases, calculate payments, and help verify returned change through voice interaction.
- Team: Saarang, Milana, Julian Kohsuke Smith
- GitHub: https://github.com/saarangpv-maker/NaviNote
- Video: https://youtu.be/-MAczQPWV6U
NextUp (Transpose (YC RFS) · Transpose (Luminal) · AWS · AGI Inc · MongoDB · EED · Bloomerang · First-Time Hacker) — MongoDB Winner#
Description: NextUp is an AI agent for volunteer coordinators. It handles cancellations, coverage gaps, plain-English requests and weather risks: it finds available, fairly-rested volunteers or alternative plans, and drafts the messages. Nothing reaches a volunteer until the coordinator approves, so the agent does the work while the coordinator stays in charge.
- Team: Woojoo Song, Woosung Lim
- GitHub: https://github.com/CosmosBoy112/next_up
- Video: https://youtu.be/J5t3U6xgwqs
NoCap (Transpose (YC RFS) · Transpose (Luminal) · AWS · AGI Inc · Bloomerang)#
Description: NoCap is a research verification platform that checks citations, source support, scientific claims, and supplied data using AI and evidence-based analysis. It flags inconsistencies, preserves an audit trail, and helps human reviewers assess academic, legal, and NGO documents.
- Team: Sri Bhargav Malluvajhula, Pranav Selagam Setty, Avishkar, Sanjay, Divyam
- GitHub: https://github.com/Bhargav072007/NOCAP-c0mpiled15
- Video: https://www.youtube.com/watch?v=L9WWBNZZF9o
OrgFlow (Transpose (Luminal) · AWS · EED · Bloomerang) — AWS 1st, Bloomerang 2nd#
Description: OrgFlow is an AI-powered knowledge management platform that helps nonprofits preserve institutional knowledge, connect documents from services like Google Drive, and turn scattered information into searchable answers, actionable procedures, and visual knowledge maps. With voice assistance, source-backed AI responses, and a volunteer coordinator agent, OrgFlow helps teams work efficiently and ensures that when someone leaves, what they know stays.
- Team: Aiden Johnson, Ryan Nair, Parth Chogule, Samarth Singh
- GitHub: https://github.com/tgmstudios/OrgFlow
- Video: https://www.youtube.com/watch?v=enjrHv29DNI
PEBO (Transpose (YC RFS) · AWS · Bloomerang)#
Description: Our product is an AI enabled personal energy building optimizer that includes a calibrated energy simulation to split an aggregate electricity bill into each building use and provide personalized recommendations to minimize electricity bills while also offering an agent (PEBO) who can answer specific questions
- Team: Yue(Zora)Liu, Kristi Maisha, JongJoo Kim, Bowen Yang
- GitHub: https://github.com/BowenYang-SBSlab/PEBO
- Video: https://youtu.be/ZTkqHJKOa6w?is=CE9aRrJh_9CbGm6s
Pizza Monsters (Transpose (YC RFS) · Transpose (Luminal))#
Description: Multiplayer AI workspace for code generation using module graph abstraction.
- Team: Mack Greenlaw-Scully, Ryan Oconnor, Jack Sangl
- GitHub: https://github.com/mgreenlaw10/compiled-15-project
- Video: https://www.youtube.com/watch?v=_Qlwu-YsHnY
Porchlight (Transpose (YC RFS) · MongoDB)#
Description: Porchlight is an AI companion that phones an older adult every day on the phone she already has. In a warm, natural voice it checks how she’s doing and actually helps her: it stops scams, books her Uber anywhere, reminds her about her pills and appointments, answers tech questions and connects her live to family. Family get a calm daily summary in the app and are alerted right away when something matters, so she stays independent and they stay in the loop without nagging.
- Team: Devank, Rishi, Manasv
Presence (Transpose (YC RFS) · AGI Inc · MICS)#
Description: Radar behind an outlet cover that tells you where people are in a room, and never who. No camera, no microphone, no new wiring.
- Team: Wyatt Schmidle
- GitHub: https://github.com/Waytts-ai/Presence
- Video: https://youtu.be/6zgb0dwHXMg
Project Spectra (Code Blooded) (Transpose (YC RFS) · Transpose (Luminal) · AGI Inc · First-Time Hacker)#
Description: A deepfake detection web app trained on over 12,000 samples of real and fake data. Runs fully locally, enhanced with Luminal Cloud.
- Team: Adam Abdel-Latif, David Bonakar
- GitHub: https://github.com/ALatif11/Project-Spectra
- Video: https://www.loom.com/share/588afd60a82b417d987ad507795031da
Rally (Transpose (YC RFS) · Transpose (Luminal) · AWS · MongoDB · Bloomerang)#
Description: A team of 8 AI bots running on Luminal Cloud works in parallel to schedule people across projects, gets measurably better every round, and takes live direction from anyone watching in plain English. On test problems the bots never saw, their schedules finish 12 to 13% sooner than the standard rule they started from, and they beat Google's industrial solver given the same 2 seconds.
- Team: Raj Mamidala, Matteo Cousineau
- GitHub: https://github.com/RM-28/Rally
- Video: https://youtu.be/Ni5PkIU0NUg
Rambl (Transpose (YC RFS))#
Description: Rambl turns a spoken ramble into a fully scheduled week — you talk like you'd talk to a friend, and it writes the tasks, finds the deadlines, and fits them around your calendar. Why schedule your tasks when you could just do them?
- Team: Rayansh Giri, Riyaz Giri, Kassidy Ke, Benson Lin
- GitHub: https://github.com/burvy/ai-todo
- Video: https://drive.google.com/file/d/1qa7pQYFui13AJnOF0QAR2UPsa9aUggx1/view?usp=sharing
RecallDB (Team Yafi) (Transpose (Luminal) · MongoDB · MICS)#
Description: RecallDB is a personal AI memory engine that transforms scattered notes, conversations, and information into structured, searchable memories. Powered by MongoDB Atlas Vector Search and Luminal DeepSeek, it lets users retrieve information using natural language and receive AI-generated answers grounded in their original sources.
- Team: Yafi Akhtar
- GitHub: https://github.com/yafiakhtar/RecallDB
- Video: https://drive.google.com/drive/folders/1IAZJBxmSKtvGHxXBX1avbNyz-qRqyYH3
Reclaimine (Transpose (YC RFS) · Transpose (Luminal) · First-Time Hacker)#
Description: Reclaimine is an AI lost-and-found for venues (airports, stadiums, concerts): it describes found items from photos so claimants must match details only the real owner knows. Parallel DeepSeek Flash agents on Luminal improve the describer round by round, and a two-judge confidence-interval gate keeps only proven gains.
- Team: Abrar Islam, Molly Barash, Younsoo Park
- GitHub: https://github.com/eeeabrar-droid/reclaimine-c0mpiled
- Video: https://github.com/eeeabrar-droid/reclaimine-c0mpiled/blob/main/video/reclaimine_v4_team_voice.mp4
ReplyDesk (ABINGTON4) (Transpose (Luminal) · MongoDB · First-Time Hacker)#
Description: Customers expect businesses to engage with their feedback, yet a 2026 study found that 71.6% of business Google reviews went unanswered, with responses taking an average of 4.2 days. Our AI-powered SaaS platform centralizes customer feedback, identifies recurring issues and potentially fake reviews, and uses self-improving AI agents to generate, evaluate, and refine personalized responses for faster, more effective customer engagement.
- Team: Maftuna Makhkamova, Naznin Jamshed Ali, Moussa Toure, Zuhan Kabeer
- GitHub: https://github.com/b-3llum/replydesk
- Video: https://youtu.be/gK_T8EALAmw
Rest (AWS)#
Description: Rest is a self-improvement app designed to help people document their journey and build habits around sleep, mental well-being, and social connection. It aims to provide a supportive space for reflection and growth.
- Team: Brian Sasaguay, Benjamin Park
- GitHub: https://github.com/GraphoLogiCode/Rest
- Video: https://www.loom.com/share/d89da0db97914b079ba9f76e05b75083
Retrofit (Transpose (YC RFS) · Transpose (Luminal) · AWS · AGI Inc · EED · Bloomerang)#
Description: Retrofit is a self-improving team of AI agents that resolves the merge conflicts companies hit when they pull upstream fixes into their long-lived forks, benchmarked on real Linux kernel vendor forks from NXP, ST and Rockchip. It borrows Luminal's search-then-measure approach: rule-based operators and parallel DeepSeek agents propose fixes, a scorer checks them against what each vendor's engineers actually shipped, and a reflector and curator turn the misses into a playbook that raised accuracy from 46% to 76% on a kernel fork it had never seen.
- Team: Madhura Khandkar, Suryansh Sijwali, Vedant Patel, Neil Barbara
- GitHub: https://github.com/SuryanshSS1011/retrofit
- Video: https://youtu.be/fBICjDVdSoE
Roster (AWS · EED · Bloomerang · MICS) — AWS 2nd#
Description: Students can't get hired, clubs can't get funding, companies can't find talent. Roster is one service that solves all three problems.
- Team: Adam Torres Encarnacion
- GitHub: https://github.com/0xABAN/c0mpiled
- Video: https://youtu.be/g0Ngz6rD-lY?si=S3DNa9XmIlRAcktf
SafeWalk Lab (404 Found) (Transpose (YC RFS) · Transpose (Luminal) · MongoDB)#
Description: SafeWalk suggests safer walking routes at night (fastest, balanced and lowest risk, from police reports and street lighting) with live GPS navigation, nearby police and blue-light phones, and live sharing protected by a code per device. In SafeWalk Lab, parallel DeepSeek agents on Luminal rewrote its routing engine to run 4.1x faster on a city they never saw, with all 500 routes identical, and every proposal and verdict is logged and analyzed in MongoDB Atlas.
- Team: Sergio Yacolca, Matthew Zhou, Justin Hnat, Evan Bonshock
- GitHub: https://github.com/Ser2104/SafeWalk-Lab
- Video: https://youtu.be/xYP59b54zus
Scam Shield (Transpose (YC RFS) · AGI Inc · MICS) — Transpose Platform 1st, AGI Inc 2nd#
Description: Scam Shield is on-device AI that listens to a phone call on speaker and warns older adults in real time when it sounds like a scam, explaining why in plain English and suggesting how to check who’s really calling. Speech-to-text and a local Llama model run entirely on the laptop, so it works with no internet and no audio ever leaves the device.
- Team: Anushka Shrivastava
- GitHub: https://github.com/itsanushkah/scam-shield
- Video: https://drive.google.com/drive/folders/1lCx-A_Ygb6mJR8M349jjeJGy7eFjMhUb?usp=sharing
SensorScout (Transpose (YC RFS))#
Description: SensorScout is an AI-inspired industrial monitoring platform that simulates machine faults and uses cooperating agents to identify blind spots, optimize sensor placement, and improve fault detection while balancing response time and cost. It features feedback-driven optimization, experiment memory, interactive visualizations, and performance benchmarking to help design more reliable sensor monitoring systems.
- Team: Sai Likhit Gadda, Aantharya Krishna
- GitHub: https://github.com/AantharyaKrishna/SensorScout
- Video: https://youtu.be/kAGnFnEJ3IY
ShiftPilot (EED · Bloomerang)#
Description: ShiftPilot is an AI agent for volunteer coordinators at student clubs and nonprofits: it fills open shifts by picking the right members from your roster and drafts a personal message to each one, plus reminders. Every draft lands in an approval queue, so the coordinator edits and approves every message before it’s sent.
- Team: Evangeline Juliet, Lamar Saleh, Ashna Thadathil
- GitHub: https://github.com/evangelinejjfk/shiftpilot_updated
- Video: https://www.youtube.com/watch?v=xKbDVaKHMww
Synth Forge (Transpose (YC RFS) · Transpose (Luminal) · AWS)#
Description: Synth Forge is an agentic AI based program that can remove AI(claude, chatgpt) watermarks from Images, text and code. It runs on local hardware or AWS cloud.
- Team: Nishant Mishra
- GitHub: https://github.com/nishantaMishra/synthForge
- Video: https://youtu.be/MAfhq3M_l2k
Tari (MongoDB · Bloomerang · MICS) — Bloomerang 3rd, MICS Winner#
Description: Tari is an iMessage-native secretary for nonprofit volunteer coordinators, backed by an agent loop that handles forms and drafts outbound messages for the coordinator to approve.
- Team: Emilia Palmer
- GitHub: https://github.com/emipalmer/coordinator-assistant
- Video: https://www.youtube.com/watch?v=LjrNd8nAadE
Threadfinder (AWS · EED)#
Description: A platform that scrapes Reddit comments and lets you reply to relevant threads.
- Team: Arta Ghasemi, Alex Filkov
- GitHub: https://github.com/artaGhs/scrapper-platform
- Video: https://youtube.com/shorts/x2odKdB15xk?is=4vHmVsvF0Cefq43p
TokenForge (Transpose (YC RFS) · Transpose (Luminal) · AGI Inc · MongoDB · MICS) — MongoDB Winner#
Description: TokenForge is a self-improving agent team: six DeepSeek-V4.1-Flash agents on Luminal rewrite the core matrix-multiply kernel of an on-device LLM (llama2.c on an M3 Pro), while a harness they can’t modify keeps only kernels that produce byte-identical output, and MongoDB Atlas vector search feeds every past win and failure into the next round. On held-out models and prompts never used during optimization, the final kernel generates text 7.82× faster than stock llama2.c (13.9 → 110 tok/s on the 110M model).
- Team: Aryaman Singh Dev
- GitHub: https://github.com/dev4-gpt/tokenforge
- Video: https://youtu.be/2ZHimN2vUzo
Training Grounds (Chess RPG) (Transpose (Luminal) · MongoDB)#
Description: Training Grounds is a Fire Emblem–style SRPG and Chess combined, where a team of parallel LLM coaches proposes battle playbooks each generation; a TypeScript engine runs thousands of sparring matches, stores results in MongoDB, and keeps the best strategy so win rate climbs over generations. We show real learning with a frozen-stats control and held-out maps/NPCs the agents never trained on.
- Team: Yoojun Kim, Jonghun Won
- GitHub: https://github.com/yjk82693/chess-rpg
- Video: https://youtu.be/3-ZxxYSwEpU
TrustLayer (ThePointers) (Transpose (YC RFS) · Transpose (Luminal) · AWS · AGI Inc · MongoDB · EED · Bloomerang · First-Time Hacker)#
Description: TrustLayer checks proposed AI actions against a business’s budget, authorization rules, and policies before they proceed. It returns an Approve, Review required, or Block decision with a clear explanation, helping businesses control AI-driven workflows.
- Team: Soham, Rhianna, Sai
- GitHub: https://github.com/rhiannaferreira/TrustLayer.git
- Video: https://youtu.be/TDf8i8FZxxg
TruthLink (thinking...) (Transpose (YC RFS) · Transpose (Luminal) · AWS · Bloomerang · First-Time Hacker)#
Description: TruthLink is an AI-powered workforce and project management tool that tracks tasks, checks deliverables through direct CAD and Simulation integrations, and learns what each team member does well. It gives project managers more time to focus on project scope and technical roadblocks by handling routine coordination and follow-ups.
- Team: George Kuney, Mitchell Goldman
- GitHub: https://github.com/mitchgold-maker/truthlink
- Video: https://youtu.be/EeUlHOptPcg
Video Workbench (Transpose (YC RFS) · AWS · AGI Inc · EED)#
Description: Video Workbench turns raw footage and text prompts into editable videos. It uses Qwen, Whisper, and JEV to analyze footage and assemble interviews, B-roll, and captions.
- Team: Alhan Akdemir, Simmy Michon, Morgan Noronha
- GitHub: https://github.com/zeroxalhan/video-production-workbench
- Video: https://youtu.be/OD7RrTLqjPc
Vitals (Transpose (YC RFS) · Transpose (Luminal) · AWS · EED · First-Time Hacker) — Engineering Entrepreneurship 3rd#
Description: Vitals scores every incoming lead for service businesses by its likely job value and chance of closing, and compares the expected work against the team's capacity to show whether the schedule will fill. It tells owners what to do next — get more leads, hold steady, or raise prices — before the schedule empties.
- Team: David Ulmer, Shanin Drucquer, Alejandro Vidal, Sergei Zhdanov, Saril Pattnaik
- GitHub: https://github.com/Flizzy50/Business-Pulse-Compiled
- Video: https://www.loom.com/share/4b4b29ffddd0447ea8022a404de2c9d2
Water Copilot (Transpose (YC RFS) · Transpose (Luminal) · MongoDB · Bloomerang)#
Description: Water Copilot is a safety assistant and memory for the operators who keep drinking water safe in remote Alaska villages. A team of agents troubleshoots problems from the plant’s own manual, escalates real health risks under rules the AI can’t override, and remembers every fix in MongoDB. Behind the scenes, self-improving DeepSeek agents on Luminal train on practice scenarios and cut errors on unseen cases by 56%.
- Team: Jia Arora, Kareen Chhangani, Malath Alassaf, Kristian Geric
- GitHub: https://github.com/karenc23/water_copilot
- Video: https://www.loom.com/share/9d69c46e4dc64c97b1def8b1437cb284
Whetstone (Transpose (Luminal) · AWS · EED)#
Description: Whetstone is a self-improving prompt optimizer: a team of AI agents running in parallel on Luminal evolves your prompts round after round and proves every gain on held-out tests it never saw. It cut errors by a third on a support-ticket classifier and tokens per run by 23% on a government contractor’s AI coding tool, and it runs as an AWS-billed subscription where scoring uses your own AI key.
- Team: Wyatt Allison, Chayce Allison
- GitHub: https://github.com/WyattA-25/Whetstone
- Video: https://www.youtube.com/watch?v=1LUy9U5utg8
Will It Material? (AGI Inc)#
Description: Will it Material. Can make anything that you want out of any material and measures it against a criteria to tell you if it's the best suit, or has a better alternative. Can also tell you what material something is made out of and suggest materials to use to create something you want.
- Team: Xena Ahluwalia
- GitHub: https://github.com/XenaAhluwalia/will-it-material
- Video: https://drive.google.com/drive/folders/1tlG2b-qukuY6bNsCBVloFM2yG85tfJ2t?usp=sharing
Workspace Zero (AWS · MongoDB · First-Time Hacker)#
Description: A Notion-style workspace with far more customization and AI-generated widgets.
- Team: Nicholas Ramirez, Benjamin Li, Akhilesh Velaga, Pavan Bharadwaj, Suhas Kothuru
- GitHub: https://github.com/SuhasProjects/WorkspaceX
- Video: https://www.youtube.com/watch?v=VZkRP0PQP_o&feature=youtu.be
YJM (Transpose (YC RFS) · Transpose (Luminal) · AGI Inc · MongoDB)#
Description: Three AI agents independently propose traffic-signal strategies, then a deterministic simulator tests them against identical traffic conditions and keeps only changes that measurably improve performance. Our self-improving agent team turns traffic control into a repeatable cycle of hypothesis, simulation, and evidence-based optimization.
- Team: Yash Patel, Jack Qin
- GitHub: https://github.com/ykp5138/Compiled-2026
- Video: https://drive.google.com/drive/folders/1Rwl5hmmG5LmwBtUaDcpI77KOFa7p5VCu?dmr=1&ec=wgc-drive-%5Bmodule%5D-goto