Skip the resume pile. Ship something real.
Get Hired Hack is a one-day, in-person hackathon at Columbia Startup Lab where NYC startups that are hiring bring real problems from their own roadmaps, and you show them what working with you looks like. The founders write the challenges, walk the floor and judge the demos. Your project is your interview.
This year's theme is AI You Can Trust. Counselors, pharmacists and investors make high-stakes decisions from messy data every day. They don't need another chatbot that sounds confident. They need AI that's grounded in real data, honest about what it doesn't know, and clear about how it got its answer.
What we're looking for: working prototypes that solve a real problem for a real company, built by engineers who care about getting it right.
What we're not looking for: slide decks with no product behind them, thin wrappers that pass a prompt to a model and print the result, or pre-built projects.
Get startedPick one of four tracks. Each one is a real problem from a startup that is hiring.
TRACK 1 | CHRP Track | Off-Grid Everything
CHRP is building a slim, modular pouch that fits behind any iPhone and turns it into an off-grid mesh radio. Phones relay messages through each other over LoRa, with no cell service and no WiFi. The hardware is their job. Your job is the software that lets it carry text, voice and video over a radio built for tiny text messages.
The constraint: each LoRa packet holds about 230 bytes. Meshtastic's fastest preset moves about 22 kbps, and its long-range presets drop to around 1 kbps or less. A one-second voice clip can take a minute to arrive, and packets get lost at every hop.
Build a messaging app that sends text, voice and video across a simulated mesh of these devices, all in software:
- Simulate the mesh: 5 or more virtual nodes on a map, each relaying packets, with realistic data rates, packet loss and range
- Squeeze the media: compress voice and video aggressively and split them into packets of about 230 bytes. Think speech codecs, tiny thumbnails, keyframes, or sending an AI-generated description and rebuilding the image on the other end
- Survive the network: acknowledgments, resending missing chunks, error correction and reassembly, so a message arrives whole even with 20% packet loss across 3 hops
- Be honest with the user: an iPhone-style app that shows exactly what has arrived, what's still in transit, the route it took and how long the rest will take. It never says "delivered" when it wasn't
What wins: a live demo that sends a voice note and an image across a 5-node simulated mesh with packet loss. Judges look at how much media you get through per byte, how well it recovers from loss, and how honest the delivery status is.
Resources:
- Meshtastic: the open-source LoRa mesh project to build on
- Meshtastic radio settings: real data rates for each preset
- LoRa mesh hardware guide (Seeed Studio)
- Meshtastic explained (Jameco)
- Codec2: an open-source speech codec that works at very low bitrates
TRACK 2 | RudCo | RFQ to Quote
Industrial buyers send requests for quotes (RFQs) as messy emails, PDFs and spreadsheets. The same item gets written five different ways ("304 SS sheet 16ga 4x8", "Stainless 304, 0.060 in, 48 x 96"), and each one needs a material grade, a spec standard, certifications and a competitive price. Today a person decodes every line by hand.
Build: An agent that takes a messy RFQ and turns it into a ready-to-send quote. It should:
- Parse each line into a standard spec (material, grade, dimensions, quantity, standard such as ASTM)
- Match each line to items in a supplier catalog
- Price each line against market benchmarks and explain why the price is competitive
- Flag compliance requirements (mill test reports, country of origin, spec certifications)
- Send any ambiguous line back to the buyer with a clarifying question instead of guessing
Requirements
Build a working prototype that addresses one of the four tracks mentioned above. It must:
- Be built during the event
- Show how it handles uncertainty: what it does when it isn't sure
- Using NicrronNinoo and Block Convey is a must in your project.
Prototypes can be web apps, CLIs, APIs with a simple front end, or notebooks with a usable interface. What matters is that it works and that a judge can check its output.
What to SubmitYour Devpost project page must include:
- Project name and one-line description
- Which track you're competing in
- What you built and why
- How your solution handles accuracy and uncertainty
- Tools and technologies used (NicrronNinoo and Block Convey is a must)
- Team member names and roles
- Link to a video demo
- Link to your GitHub repository
Prizes
Interview Slot
Get interviewed for a startup.
Judges
Somya Gupta
Judging Criteria
-
Technical Execution (5)
Does it actually work, and does it get the right answer? -
Trust & Explainability (5)
Can a user see why the AI said what it said? Does it flag what it doesn't know instead of bluffing? -
Design (5)
Could a counselor, pharmacist or investor use this tomorrow without a walkthrough? -
Uniqueness of Idea (5)
Did you find a sharper angle on the problem?
Questions? Email the hackathon manager