GDG Coding Jams
Build with AI. Together.
Share
GDG Brooklyn · United States

AI Builder Day at NYU School of Engineering

This week you’re building
💡
Your Own Idea

No menu. No template. The thing you've been daydreaming about.

Full brief for Track 09
What ships today

The version everyone leaves with.

Bring an idea you've been sitting on — write a 1-paragraph PRD, hand it to Antigravity, and ship a working slice in 45 minutes. No starter repo. No menu. Just your spec and your taste.

While you build

Things worth deciding early.

  • 1One paragraph of PRD beats a backlog. If you can't describe it in three sentences, you can't ship it in 45 minutes.
  • 2Pick the smallest version of the idea that's still recognizable — one core flow, one screen, no settings.
  • 3Your signature detail matters more here than anywhere else. There's no demo to copy from, so the soul has to come from you.
  • 4Stuck on prompting? Steal from the other tracks — image gen, RAG, persona design, agent loops are all fair game.
The polished version

What the at-home build adds.

  • Whatever the at-home version of your idea looks like.
  • Steal patterns from tracks 1-8 — they're reference implementations now.
  • Polish pass: empty states, error states, the one delightful detail.
The shape of the session

Two hours, four movements.

  1. 0:00 – 0:15
    Intro

    Grab a name tag and a slice, say hello to the instructor, and meet the people either side of you — you'll be asking them things later, so it pays to know their names now. Low pressure, high creativity, ship something messy. Then a quick word on what the room is building tonight, and you're off.

  2. 0:15 – 0:30
    Credits & setup

    The instructor shares the Google Cloud credits with the room. Everyone gets their environment ready — whichever tool they brought. Nobody should be installing anything once the build starts.

  3. 0:30 – 1:35
    Build

    Heads down, and enjoy it — this is the part everyone came for. The agent does the typing; you decide what it's making. Ask the room when you get stuck; someone two seats away hit the same wall ten minutes ago. Want more structure? The codelab walks the spec-driven route — Setup → Plan → Review → Build → API → Verify — but nobody has to follow it.

  4. 1:35 – 2:00
    Share & submit

    Show the room what you made — two minutes, screen-shared from your seat. Half-finished is welcome; the wobbly ones are usually the most interesting. Say what surprised you along the way, and cheer for everyone else while you're at it. Then submit your build before you leave. This is the one thing not to skip — it's what puts your name and your chapter on the showcase, and it takes about two minutes.

Shipped from this jam

52 builds.

Add yours →
Anatomy Atelier
Track 09Universiti Tun Hussein Onn Malaysia
Anatomy Atelier

Anatomy Atelier is an interactive 3D human anatomy and biology learning platform built for medical students, educators, and SPM Biology candidates. It combines real-time WebGL 3D organ specimens with an intelligent RAG search engine powered by Gemini API to query official syllabus textbooks. Learners can isolate organ tissues, simulate biological metrics, and ask physiological questions with instant, verified textbook citations.

I expected RAG textbook ingestion and 3D WebGL rendering to create heavy latency, but pairing Gemini API with structured retrieval delivered instant, context-aware anatomical answers while maintaining a smooth 60fps canvas.

Gemini APIGoogle AI StudioAntigravityCloud Run
SensePath
Track 09GDG Brooklyn
SensePath

SensePath is a voice-first accessibility assistant designed for blind and low-vision users. It combines Gemini Live, multimodal vision, and Google Maps Platform to help people navigate the last few meters of a journey—finding entrances, benches, elevators, and understanding their surroundings through natural conversation. The standout feature is its ability to seamlessly switch between navigation, object finding, and immersive scene descriptions while keeping the interaction hands-free and acces

We expected navigation to be the hardest problem, but the real challenge was making AI know when not to speak. Designing an assistant that gives timely, concise guidance without overwhelming or distracting the user turned out to be just as important as the underlying AI itself.

HarmonicBlend
Track 09GDG Brooklyn
HarmonicBlend

HarmonicBlend is an AI-powered, browser-based DJ and stem mashup workstation that allows users to search any song on YouTube Music and turn it into a professional-grade remix in real time. The platform features dual interactive vinyl turntables, scrubbable waveform timelines, 3-band EQs, lowpass/highpass filter sweeps, crossfader controls, and an offline WAV exporter that lets users render and download their completed mixes directly to disk, as well as using Facebook's Demucs ML to create stems.

Using Gemini within Google Antigravity to build this project was a leap in autonomous, agentic software engineering. Its crazy how Antigravity reasoned across full-stack boundaries, making Python Flask backends, setting up PyTorch ML pipelines, and coding complex Web Audio API DSP frontend engines, something I would need months of training and understanding to build.

Drop
Track 09GDG Brooklyn
Drop

Drop is a mobile-first web app that turns New York City into a live board game where physical items like disposable cameras and toys drift from stranger to stranger. Built for urban explorers and romantic wanderers, the app lets you drop and pick up items and monitor a drop's entire journey across the city while staying completely anonymous.

Google AI Studio is a surprisingly friendly entry point for non-developers, proving that you don’t need a traditional coding background to bring a complex spatial and interactive idea to life.

Drift City
Track 09GDG Brooklyn
Drift City

Drift City is a community-driven web app that turns everyday urban spaces into an ongoing game of real-world object pass-along. Built for urban explorers, scavengers, and curious city dwellers, it tracks physical items—like a disposable camera or a rubber duck—as strangers move them from corner to corner across New York City. The cool part? You get to watch real-world human connections trace interactive, living vector paths across a digital map in real time.

Letting Gemini rephrase my prompts for AI Studio was helpful.

Nomad-Quest Game
Track 09GDG Brooklyn
Nomad-Quest Game

Nomad Quest is an interactive travel etiquette and cultural survival simulator. Instead of reading dry travel blogs or memorizing standard phrases, users are dropped into immersive "24-hour" city scenarios where they must navigate real-world dilemmas, cultural nuances, and language barriers. It gamifies global citizenship—rewarding players with passport stamps and respect points as they learn how to navigate new cultures without making costly, awkward tourist mistakes.

The "Homework" Trap: It was surprising how quickly educational content can accidentally feel like a school assignment (the "Quizlet effect"). We realized that simply having good information isn't enough; the delivery has to be dynamic. By introducing bouncy, engaging animations and restyling the scenarios to feel like an adventure game rather than a multiple-choice test, the exact same educational

Nestwork
Track 09GDG Brooklyn
Nestwork

Nestwork matches small groups of compatible strangers in NYC - by shared interests like tech, wellness, trekking, or being new to the city - and sends them a curated invitation to a real evening out, itinerary and venues included. Every member is ID-verified before their first Nest, so meeting strangers actually feels safe. Once you RSVP, a scoped group chat opens just for your Nest, active through the event and archived after - no lingering inbox, no swiping.

We assumed the hard part would be the matching algorithm. It wasn't - the hard part was realizing trust, not compatibility, was the real blocker to strangers meeting up. The ID verification step ended up being the feature people cared about most.

ForgeIQ: Multimodal Engineering Intelligence
Track 09GDG Brooklyn
ForgeIQ: Multimodal Engineering Intelligence

ForgeIQ is a multimodal engineering intelligence platform powered by Gemini that analyzes technical drawings and component images. Users upload engineering artifacts and receive structured AI-assisted reviews including design analysis, manufacturability (DFM), material recommendations, preliminary FMEA, quality inspection checklists, and prioritized engineering recommendations. ForgeIQ helps engineers evaluate designs faster while keeping the human engineer in the decision loop.

I was surprised by how much structured engineering insight Gemini could generate from a single technical drawing. The biggest challenge was designing prompts that produced engineering-focused analyses rather than generic image descriptions.

scout
Track 09GDG Brooklyn
scout

Scout is an AI apartment-hunting agent for NYC renters. It turns plain-English preferences into listing searches, ranks and enriches the matches, and organizes next steps in one map-first tracker. It can also draft broker outreach so renters can move quickly while staying in control of what gets sent.

I expected the AI to be most useful for writing messages, but its biggest value was translating messy human preferences into precise search filters. Once Scout understood requests like “a no-fee two-bedroom under $5,000,” the rest of the workflow became dramatically simpler.

Food Hunting
Track 09GDG Brooklyn
Food Hunting

Food Hunt allows you to pick your food when you don't know what to eat for lunch/dinner. The website will prompt you to enter basic information such as your address, the radius of the website, and your budget. Then it will ask you select the kind of food that you want right now.

It was a very tough project to do within a short amount of time. But we managed to finish on time

TeamPulse Project Manager
Track 09GDG Brooklyn
TeamPulse Project Manager

TeamPulse streamlines team collaboration by automatically assigning tasks to the most suitable team members. It allows teams to define each member's role, skills, and weekly availability, then intelligently matches unassigned tasks based on expertise and capacity. This helps teams, especially large or remote ones, delegate work more efficiently and complete projects more effectively.

What surprised me most during development was how seamlessly role-based AI task delegation transforms project planning. Rather than just assigning tasks randomly, evaluating member skills, weekly capacity hours, and target deadlines using Gemini AI produces surprisingly nuanced, practical workload recommendations.

IBM Escape Room
Track 09GDG Brooklyn
IBM Escape Room

This project is an Autonomous AI Escape Room Engine that leverages Yann LeCun’s JEPA architecture and Gemini 2.5 Flash to create procedurally generated 3D games. It utilizes a multi-agent framework where specialized modules like the Configurator and World Model collaborate to simulate the physics and logic of your environment. Players select unique themes and scenarios which the engine then transforms into an interactive, real-time 3D world using the Three.js library. The system moves beyond sim

Mapping Yann LeCun’s JEPA paper directly to game master mechanics revealed that modular brain agents (Configurator, World Model, Critic, Actor) naturally eliminate AI hallucinations by simulating spatial physics before narrative generation. We were surprised by how seamlessly an LLM orchestrator could generate real-time 3D spatial JSON protocols while maintaining strict object permanence across ga

Pocket Maps
Track 09GDG Brooklyn
Pocket Maps

Pocket Map is a mood-driven location discovery app where users snap and pin cozy hidden gems to a shared map. Guided by Pickles, an empathetic AI raccoon companion, users describe how they feel—and Pickles analyzes community photos to match their exact vibe with the perfect local sanctuary.

We are really surprised by the capabilities and speed of Google AI Studio.

Velocity Neon 3D
Track 09GDG Brooklyn
Velocity Neon 3D

The project involved creating an immersive 3D racing game set in a captivating sci-fi universe. Players find themselves in a futuristic landscape filled with neon-lit cityscapes and breathtaking alien terrains.

Google Pro is an advanced and versatile modeling tool that allows users to quickly prototype and craft immersive 3D environments. With its intuitive interface and robust features, you can seamlessly bring your creative visions to life, transforming ideas into vibrant digital landscapes almost instantly.

Vibe Style
Track 01GDG Brooklyn
Vibe Style

This app is a conversational AI stylist for your closet: it asks about your mood, who you’re dressing for, and the occasion (or trip details), then generates complete, head-to-toe outfits pulled from your own digitized wardrobe — shown as e-commerce-style look cards with Add to Bag, Favorite, and Customize options. It also handles multi-person trip packing (partner/family), weather-aware travel essentials, and destination-only shopping suggestions, all wrapped in a friendly “best friend stylist”

I assumed the hard part would be the styling logic. It wasn’t, the hard part was getting the AI to remember it already knew you were shopping solo, and to stop asking.

VibeTrav
Track 09GDG Brooklyn
VibeTrav

VibeTrav is a medical tourism platform that helps people discover and plan authentic healthcare journeys, with a dedicated focus on Ayurveda. The new Ayurveda Centers feature lets users explore highly recommended treatment centers first, then browse a wider directory with powerful filters by location, treatment type, and budget. It is designed for international patients seeking trusted, high-quality Ayurvedic care without the usual research overwhelm.

What surprised me most was how strongly users wanted a clear separation between “highly recommended” centers and the broader list. Prioritizing quality over quantity proved more important than simply showing a large number of options.

TourCare
Track 09GDG Brooklyn
TourCare

TourCare NYC is an AI-powered travel assistant designed to help visitors and residents explore New York City through personalized, real-time recommendations. The solution combines conversational AI, location-aware intelligence, and personalized itinerary generation to create a seamless travel experience. Users can ask natural language questions such as: "What are the best family-friendly attractions near me?" "Plan a half-day itinerary in Manhattan." "Suggest accessible attractions within 30 m

Team work, AI capabilities, Real word problem

Side Quest NYC
Track 09GDG Brooklyn
Side Quest NYC

SideQuest NYC is a personalized event discovery app for people who want spontaneous, local things to do without digging through dozens of websites. Users can search by voice or filters like interest, budget, location, and vibe to find nearby events happening now or in the next few days. The cool part is that it understands requests like “find something low-key and not too crowded” and turns them into tailored recommendations.

One of the biggest challenges was accurately scraping event locations and times from multiple websites, since each source formatted its data differently. Normalizing inconsistent dates, addresses, and missing information made it difficult to produce reliable real-time recommendations.

Art Attack
Track 09GDG Brooklyn
Art Attack

Art Attack is a real-time two-player fighting game where each player draws their own character, then AI turns the sketch into a fighter with stats, abilities, a sprite, and an arena theme. Players can join from another device, lock in their characters, watch the battle intro, and fight live in the same arena. The cool part is that every match begins with something completely original that the players created themselves.

I was surprised by how many small real-time state changes have to stay perfectly synced for the experience to feel seamless. A single missed update after character lock-in could prevent the entire match from reaching the battlefield, which made reliable multiplayer flow just as important as the AI character generation.

SafeLink
Track 09GDG Brooklyn
SafeLink

SafeLink is a one-tap emergency app for anyone who feels unsafe on the street or gets caught in an accident. It starts recording and tracking your location the moment you open it, then lets you send that location and footage to police, ambulance, or fire with a single button — or dial 911 directly. Gemini turns whatever you can manage to type into a structured dispatch report, so responders get clear information even when you can't explain the situation yourself.

How much of "send a live stream to the police" turns out to be impossible in a browser. There's no endpoint to send to, and phones won't even record both cameras at once — so most of the work became figuring out what was real versus what I had to honestly label as simulated.

Film Room
Track 01GDG Brooklyn
Film Room

Film Room turns a phone video of one pickup basketball possession into a coach's breakdown — annotated video, real spatial analytics, and a bridge to the NBA action it resembles — so you learn to read the game by studying your own film.

The model turned out to be far better at reading basketball than at telling my friends apart — it diagnosed the blown switch instantly but couldn't reliably say who blew it, which forced me to design the whole system so the tactical read still holds when player identification fails.

Air Studio
Track 09GDG Brooklyn
Air Studio

Our project is a multiplayer AI-powered rhythm game where two players compete by playing virtual instruments using hand gestures in front of a webcam. The game tracks each player's movements in real time and scores their performance based on accuracy and timing. It creates an interactive and fun way to enjoy music without needing any physical instruments.

We were surprised by how challenging it was to keep the multiplayer gameplay smooth while accurately tracking hand gestures in real time. Even small delays could affect the overall gaming experience.

CandidAI
Track 09GDG Brooklyn
CandidAI

Remote interviews have a trust problem right now: candidates using invisible AI tools to feed them answers live, and in some cases, people who aren't even who they say they are. CandidAI is a multi-agent system that sits in the interview, watches for those red flags across identity, behavior, and language, and quietly tells the interviewer, "hey, might want to ask a follow-up here." It never rejects anyone; it just makes sure the human making the call actually has the full picture.

Honestly, the detection part was the easy half. What actually kept me up was the guardrail design, figuring out how to flag real cheating without accidentally penalizing someone who's neurodivergent, has bad wifi, or talks with a different rhythm than what the model expects as "normal."

Napkin
Track 09GDG Brooklyn
Napkin

Napkin turns photographs of hand-drawn app sketches or whiteboard mockups into live, interactive web applications in seconds. Powered by Gemini Vision and code generation, Napkin extracts UI components, pre-populates stateful demo data, and instantly serves a shareable web app complete with device preview frames and QR codes for mobile testing. It bridges the gap between raw paper brainstorming and functional software prototyping effortlessly.

What surprised us most was how gemini transformed pen-and-paper scribbles into fully web applications without requiring design templates.

Daily Standup
Track 09GDG Brooklyn
Daily Standup

Most people quit health tracking because apps make it feel like a chore. Daily Standup lets you log your entire day by sending one text to a Telegram bot. AI parses it instantly and populates a personal dashboard with trends and charts. Thirty seconds a night, zero forms.

The macro bug surprised me most when dinner overwrote lunch, it looked like a Gemini extraction failure, but Gemini was actually right both times. The bug was in the application logic merging the results, which is a good reminder that AI output is usually fine; it's the code around it that breaks things.

snap311
Track 09GDG Brooklyn
snap311

Snap311 helps New Yorkers report visible street and public-space issues with almost no manual work. Users take or upload a photo, and Gemini identifies the problem, uses the photo or device location, and checks real NYC 311 data for open duplicate reports nearby. If it’s new, Snap311 creates a clear report for the user to verify and simulates submitting it.

I learned that Gemini is much more reliable when its image analysis is constrained to a short list of real NYC 311 complaint types instead of being asked to freely label anything it sees. I also learned that surfacing duplicate reports early is a UX win: it avoids making users fill out a full form only to discover the issue was already reported, so I fixed the flow by checking nearby open reports.

shinobi-seals
Track 09GDG Brooklyn
shinobi-seals

Pokémon Go, but you're the character. Using your phone's front camera, we read real hand signs in real time and turn them into attacks or defense, no buttons, just your hands. Health is permanent, chakra regenerates and fuels your moves, and health hitting zero means you lose. Find a nearby player and battle it out live, face to face.

I assumed relying entirely on the front camera for real-time hand signs would feel unresponsive without physical buttons. Turns out that removing the on-screen UI makes the face-to-face PvP battles surprisingly intense and deeply immersive.

Host on google ai studio via google web endpoint The https://8thwall.org
Track 09GDG Brooklyn
Host on google ai studio via google web endpoint The https://8thwall.org

Host on google ai studio via google web endpoint of dead 3d AR from 8th wall The https://8thwall.org then built w ai studio web url https://ai.studio/apps/36efe82a-0b9b-4683-b73f- https://ais-pre-omiblay4dw4a5un3kntekd-638685251003.us-east1.run.app ai studio link off his game description off Claud https://docs.google.com/document/d/1UJglgOpu0i4Q5QhPi_BDjejfb51wvkaxZc2Fys3JYO4/edit?usp=drivesdk & game https://aistudio.google.com/apps/3ad79437-f262-4667-acc0-6d1ce6138b03 & Ai Studio app off it

I am using the claude Project definition doc

Previs
Track 06GDG Brooklyn
Previs

Magic Hour turns a screenplay into a consistent storyboard: your characters keep the same face and voice from scene 1 to scene 18, because every generated frame is scored against a locked identity card. A pre‑viz studio for filmmakers, with AI shot lists, in‑character dialogue, and real location scouting.

I assumed character drift was the image model being inconsistent but the model was fine all along; the drift came entirely from us re‑describing the character on every call instead of compiling her look once and reusing it verbatim.

NYC Point-to-Point
Track 09GDG Brooklyn
NYC Point-to-Point

The NYC Point-to-Point Travel Recommender is a smart, interactive routing engine designed for New Yorkers and tourists who need help navigating the city's complex transit web. Instead of just showing the fastest route, it calculates and ranks all available options—from the subway to rideshares to Citi Bikes—based on your live priorities like cost, safety, and group size. The coolest part is the dynamic scoring engine: as you drag the priority sliders or increase your group size, the app instantl

I was surprised by how dramatically the math shifts when factoring in group size; a seemingly expensive $40 rideshare instantly becomes the most cost-effective and convenient option when split among four people, entirely upending the standard "the subway is always cheapest" assumption.

Inclusivity AI
Track 09GDG Brooklyn
Inclusivity AI

Inclusivity AI is your friendly helper for everyday life. We make scheduling, shopping, and budgeting simple and accessible, with gentle reminders along the way! Made simple for people with disabilities, memory challenges, or anyone who just needs a little extra support.

Google Studios is fun to use and you can customize certain parts of your app in minutes! Also special shoutout to Milan for helping me understand how to upload it to Github!

SensePath
Track 09GDG Brooklyn
SensePath

SensePath is a voice-first accessibility assistant designed for blind and low-vision users. It combines Gemini Live, multimodal vision, and Google Maps Platform to help people navigate the last few meters of a journey finding entrances, benches, elevators, and understanding their surroundings through natural conversation. The standout feature is its ability to seamlessly switch between navigation, object finding, and immersive scene descriptions while keeping the interaction hands-free and acces

We expected navigation to be the hardest problem, but the real challenge was making AI know when not to speak. Designing an assistant that gives timely, concise guidance without overwhelming or distracting the user turned out to be just as important as the underlying AI itself.

Rainy Day Rhythm
Track 09GDG Brooklyn
Rainy Day Rhythm

Generate a beat map to compliment any song you want to turn into a rhythm game. Connected through Youtube, you can either insert a link or describe the video you are looking for. There are additional features to make it more entertaining.

I was surprised the AI filled in so may gaps in what i didn't tell it to do. For example, I intended for it to be played through both keyboard and touch screen, but I didn't specify how to do a flick note using key binds. The AI filled it in my equating double-tap as a flick note.

Song Battle
Track 09GDG Brooklyn
Song Battle

Song Battle PK turns any artist, era, or musical mood into a fast head-to-head song tournament. Players watch YouTube performances, choose winners, and crown a champion—or generate an original song and synchronized AI music video from their own creative prompt.

the sound genertaing was way better than I thought

NYCity SOS
Track 09GDG Brooklyn
NYCity SOS

City SOS is a voice-first safety companion that helps people ask for assistance even when they can't speak openly. It starts as a normal NYC and campus resource assistant, then seamlessly switches into a natural cover conversation (like ordering pizza) while quietly gathering context. When the user confirms with a predefined trigger phrase, it discreetly alerts their pre-selected trusted contacts without interrupting the ongoing conversation.

The biggest surprise was how convincing the cover conversation felt. With Gemini Live and careful prompting, the assistant stayed completely in character while still collecting the context needed for a discreet alert.

Handkerchief Duel
Track 09GDG Brooklyn
Handkerchief Duel

Developed a web-based strategic handkerchief game built on a 10-node number line and finite-state machine, implementing real-time two-player gameplay with Socket.io and AI opponents. The game features asymmetric roles—a Dropper strategically decides whether and where to drop the handkerchief, while a Guesser chooses whether to open their eyes to infer the drop, with damage determined by the handkerchief's placement and each player's decisions.

I was surprised by how capable AI Studio is—it generated most of the application with very little code from me. The only thing I missed was the ability to force or replay specific scenarios, which would make manual testing much easier.

Class Scheduler
Track 09GDG Brooklyn
Class Scheduler

My project can create a course schedule for students based on a course list it can draw from as well as an option for required and optional courses. It takes into account professor rating, how close together are the classes, and preferred days off giving multiple schedules with a score for each based on how the user weighted each factor.

What surprised me was how AI seemed to think ahead as it added features that alongside the ones, I suggested to make my suggestions work better.

ink2pix
Track 09GDG Brooklyn
ink2pix

Ink2Pix is an interactive creative canvas that transforms hand-drawn sketches into hyper-realistic 2K images and animated video clips in real time. Designed for visual storytellers, artists, and designers, it turns rough doodles into high-fidelity artwork using customizable style presets. The cool part? With a single click, Gemini Omni Flash animates your static artwork into dynamic 3D video scenes complete with synchronized ambient audio.

Gemini could handle some edge cases we didn't expect. We didn't expect such a great output for the prompt we gave.

contest_It
Track 09GDG Brooklyn
contest_It

Contest It is a civic AI tool that tells NYC residents whether they should fight or pay a city summons in seconds. Upload a photo of your ticket (or type the violation code), and Gemini Vision OCR extracts the charge, cross-references it against 421,000+ real OATH hearing outcomes, and shows your actual dismissal probability with a ready-to-submit defense brief. It's built for the everyday New Yorker who gets a $300 sidewalk ticket and has no idea that historically, 45% of those get thrown out

The biggest surprise was the data itself we found that for some of the most common NYC violations (noise, sidewalk snow, building maintenance), dismissal rates are shockingly high, yet almost nobody contests them because the process feels intimidating. The gap between "you'd probably win" and "people just pay" turned out to be the entire reason this project matters. On the technical side, we did

Gotham Intel
Track 09GDG Brooklyn
Gotham Intel

GothamIntel is an NYC sublet marketplace where seekers swipe and search for homes, and hosts manage listings, inquiries, and messages. Gemini powers vibe search (text, image, or voice), a listing chat assistant, and neighborhood intel; safety, noise, transit tips, and lease takes grounded in local listing data.

We were surprised how quickly we could build with AI Studio.

Voice Arena
Track 09GDG Brooklyn
Voice Arena

Where physical brawl meets vocal brawl! Voice Clash: Smash Arena is an innovative local 2-player platform fighter that transforms your voice into your ultimate weapon. Players maneuver using fluid keyboard controls while unleashing powerful elemental attacks and cinematic combos simply by shouting skill words into a single shared microphone — completely hands-free!

Gemini got the gist of the game in the first go, so I get to spend rest of the time making small adjustments such as card positioning and adding player camera.

Resil-AI
Track 09GDG Brooklyn
Resil-AI

ResilAI is an AI Incident Readiness Platform that helps enterprise security teams automatically measure and remediate AI-specific risks. Instead of relying on manual questionnaires, it ingests live telemetry from SIEMs (like Splunk and Wazuh) to calculate a deterministic Governance Health Index. The coolest part is our Dual-Engine Architecture: we use hard, verifiable math to calculate the risk scores, and use Google Gemini for the executive summary.

I was most surprised by how perfectly we could use Gemini to translate highly technical, deterministic telemetry data into clear, executive-level narratives without any hallucination. By strictly separating the "math" (our scoring engine) from the "speech" (Gemini), we proved that AI can make compliance ledgers highly readable while remaining 100% legally auditable.

911OperatorAI
Track 09GDG Brooklyn
911OperatorAI

Our project is an AI-powered, multilingual emergency triage platform that eliminates language barriers in crisis situations. When a user calls in their native language, our Twilio and Gemini-powered backend instantly transcribes, translates, and extracts their exact emergency and location. This data is streamed in real-time to an operator dashboard, which maps the caller's location alongside the nearest police stations and fire departments for rapid, life-saving dispatch.

Building this highlighted the incredible potential of combining real-time telecom APIs with generative AI to solve life-critical communication barriers. Successfully orchestrating Twilio, Gemini, and Supabase into a live pipeline was challenging, but seeing a foreign-language phone call turn into a pinpointed map rescue route made it all worth it.

RESERVED NYC
Track 09GDG Brooklyn
RESERVED NYC

Don’t know where to eat in New York City? RESERVED NYC helps solve this problem by finding restaurants that are perfect for you based on your cuisine preferences, the occasion, the vibe you are looking for, the location, and whatever else you can think of. We want to make sure you find that perfect restaurant for you! You can also generate custom invitations to send to your guests, create custom restaurant guides, and see where other people are eating too.

Initially, we did not put much thought into the prompt. But as we were working on our app, there were changes we wanted to make that were harder to implement. What we learned was that it’s very important to design a good prompt to make things work

NYC Affordable Housing & FRFP Portal
Track 09GDG Brooklyn
NYC Affordable Housing & FRFP Portal

A targeted rent freeze program for households earning 30–60% of Area Median Income, combined with landlord subsidies and aggressive affordable housing production. This two-pronged approach stabilizes existing tenants while addressing long-term supply constraints.

how easy it was to use the google AI ecosystem

AI Interviewer
Track 09GDG Brooklyn
AI Interviewer

AI Interviewer is a real-time, voice-enabled mock interview platform that helps software engineers master both algorithmic coding and system design interviews. Built for tech candidates seeking realistic practice without the hassle of scheduling, it pairs an interactive code editor and architectural whiteboard with customizable interviewer personas. Powered by the low-latency Gemini voice AI that dynamically probes edge cases mid-interview and delivers comprehensive performance breakdown after.

One of our challenges was to stop Gemini from giving away solutions too early. What surprised us was how naturally it calibrated its guidance, offering subtle nudges while staying in character according to the selected persona, acting as a real interviewer.

Twin Space Dodger
Track 09GDG Brooklyn
Twin Space Dodger

Twin Space Dodger is a retro-styled dual-ship space runner designed to test parallel motor control and split-brain reflexes. Players control two ships simultaneously across independent parallel lanes to dodge incoming asteroid storms. With zero shooting, custom Web Audio synthesis, desktop dual-keyboard mapping (WASD + Arrows/IJKL), and mobile landscape dual-thumb touch zones, difficulty ramps up smoothly after a 5-second per-lane warmup into full dual-sync chaos.

I was surprised by how effectively the 5-second single-lane warmup trained players' left-and-right hand dexterity before launching them into dual-sector chaos. Fine-tuning the Web Audio synth and parallel collision detection made the retro arcade feel remarkably responsive and addictive.