A multi-chapter regional initiative uniting chapters around annually rotating societal challenges to build solutions with Google's applied AI ecosystem.
The GDG AI for Good initiative is a multi-chapter regional program across the Google Developer Groups in North America. Rather than running disconnected, one-off hackathons, the initiative unites chapters around an annually rotating societal challenge (such as environmental resilience, digital accessibility, public health informatics, or workforce readiness).
Local chapters host hacking sprints during the Fall period, where developers, designers, students, and domain experts collaborate to build solutions for community-validated problems using Google’s applied AI ecosystem.
Raise awareness of societal challenges and educate the public on the latest AI tools.
Provide a concrete on-ramp for participants to master agent frameworks within AI in an applied, ethical problem space.
Establish a shared evaluation baseline across chapters, fostering cross-chapter collaboration, talent discovery, and measurable impact tracking.
Builds a direct bridge between local non-profits, municipal agencies, and GDG talent, delivering actionable open-source repositories to underserved community partners.
Connects collegiate tech enthusiasts, career switchers, and senior engineers across urban and suburban tech corridors, showcasing regional talent to industry sponsors and ecosystem partners.
This proposed 4-pillar rubric serves as the master scoring guidelines across all chapters and annual iterations to guarantee cross-regional parity.
| Pillar | Weight | Core Focus | Scoring Criteria |
|---|---|---|---|
| 1. Problem Validation & Social Impact | 25 Pts | Depth of community relevance, beneficiary focus, and quantifiable real-world value. |
|
| 2. Technical Innovation & AI Architecture | 25 Pts | Rigor, elegance, and utility of the underlying solution. |
|
| 3. Responsible AI, Ethics & Accessibility | 25 Pts | Safety, equity, bias prevention, transparency, and inclusive design principles. |
|
| 4. Feasibility, Scalability & Sustainability | 25 Pts | Viability of handoff, operational cost management, and long-term ecosystem maintenance. |
|
Participants retain 100% ownership of all software, models, and intellectual property created during the event; neither Google Developer Groups (GDG), Google LLC, nor host institutions claim any equity or commercial rights in your work. To ensure community and non-profit partners can deploy, maintain, and scale these civic solutions without legal barriers or licensing friction, all builds submitted to the GDG AI for Good track must be released publicly under the Apache License 2.0, with accompanying project documentation and synthetic evaluation datasets shared under Creative Commons (CC-BY 4.0 or CC0).
By submitting to codingjam.dev, teams warrant that their build is original work, free of unauthorized third-party or employer-owned proprietary trade secrets, contains no unscrubbed personally identifiable information (PII), and adheres strictly to upstream AI foundation model terms of service. Entrants grant GDG and local organizing chapters a non-exclusive, perpetual, royalty-free license solely to index, demonstrate, screenshot, and publicize the project across regional leaderboards, promotional media, and DevFest showcases.
Neither is the beginner option. It comes down to what's already on your laptop, and whether you want a desktop IDE at all. Four questions, then bring the one that fits.
Already know which one you want?
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.
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.
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.
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.
Our project is called Home Lens and its a way for renters to avoid unfair deductions from landlords. Renters scan any letters along with their house, and our app matches it with the VA law to find any unfair deductions from the security deposit.
“I was surprised with Gemma's capabilities to analyze damages during move in from our 3d scan and its comparison to Virginia Law for unfair deductions from the security deposit.”
ADAPT is a tool that helps people make their homes easier to move around in, especially people with mobility issues. Users can upload a floor plan, explain what they need, and enter their budget. ADAPT then finds areas that could be improved and suggests changes. It also shows the before and after in 2D and 3D so users can see what their home could look like.
“What surprised us was how we were able to connect different technologies to analyze an uploaded floor plan. We had to get the frontend, backend, AI, and geometry system to work together to identify problems and suggest changes. It was cool to see how all these different parts came together to make our idea work.”
Clearlease is a pre-sign check for renters in Virginia, DC and Maryland. Paste a rental listing or snap a photo of a lease, and Gemma 4 reads it into facts with exact quotes; our code verifies every quote and decides three verdicts: is the price normal (HUD and Census data), are there FTC rental-scam warning signs, and which lease clauses conflict with state law. Every flag shows the words that triggered it, and nothing you paste is stored.
“A simple keyword filter was about as good as our AI on ordinary listings; Gemma only earned its keep on the disguised scams (misspellings, Spanish, screenshots, and "we never take Zelle" disclaimers). That changed our pitch from "AI catches scams" to "AI catches the scams keywords miss".”
HomePick DMV is a web app that helps first-time renters in the DC, Maryland, and Virginia area find income-restricted apartments. You enter your income, then pick between two curated buildings at a time, and it learns what you care about. In the end, it hands you a final short list of homes with photos, what you’d actually pay, who to call, and much more. It also reads rental applications with Gemma 4, turns them into a simple checklist, and warns you about scams.
“We assumed affordable housing was a niche field, with only a few homes available. To our surprise, about 122,000 income-restricted homes across the DMV sit in public records that almost nobody reads.”
Roomwise is a housing research tool for Virginia Tech students that shows what a listing really costs. It separates per-person, whole-unit, monthly and installment prices, and flags the fees and lease facts still missing. Its cool part is a local, fine-tuned Gemma 4 assistant that explains the evidence, while the app's code keeps every price and calculation exact.
“I did not expect the google search engine to be a seperate feature not present inside the google cloud console”
Our tool helps renters document the condition of their rental so that they can protect their security deposit. A renter uploads a matched move-in and move-out photo of a room, and Gemma 4 (through the Gemini API) flags possible differences between them. The AI only suggests, as the renter reviews each flagged item, confirms or dismisses it, and adds a note. It only uses the confirmed items from the renter. From those items, the app makes a PDF evidence report with side by side photos, photo time
“How Gemma AI connected with other tools such as streamlit, pillow, etc allowed us to make a functioning website with actual use in only a day. Despite it being a rough submission, this is something that can be scalable and actually useful.”
For renters, especially college students signing their first lease, reporting a maintenance issue can be frustrating: taking photos, describing the damage, writing emails to landlords, and keeping track of previous complaints, all while worrying that their concerns may be ignored or disputed. FixProof simplifies this process by using Google's Gemma 4 AI to analyze repair photos, generate editable maintenance reports, and draft professional landlord emails in seconds. Every report is timestamped
“What surprised us most was how much more difficult it was to make AI trustworthy than to make it functional. Getting Gemma 4 to analyze repair photos was only the beginning; the real challenge was building safeguards around its uncertainty, protecting renters’ data, and creating verifiable records that people could confidently review and control.”
GAURD is an all in one tool that helps potential house buyers run a security check on their prospective city to ensure they are well informed about potential risks and crime rates. This security check not only provides current information about both violent and property crime, but also has a trained gradient boosted trees algorithm which predicts whether crime will rise in a given city 4 years from now.
“I excepted the promoting of Gemma to be simpler while using an API, but getting it to deliver standardized and accurate responses was difficult and took a mix of back end design changes and prompt engineering. I was impressed with the accuracy of our self made LightGBM when it came to predicting changes in crime rate as there were few studies on using Machine Learning to predict crime rates”
TerraSim tackles the housing shortage by exploring development opportunities on available or underused land. Outline a lot on a 3D map and describe your ideal housing project. Using Gemma via the Gemini API, TerraSim generates site-specific floor plans, furnished 3D walkthroughs, material quantities, and estimated costs. These estimates support early planning but do not replace professional engineering, code reviews, contractor quotes, or formal feasibility studies.
“We initially assumed that Gemma 4 was not that capable, and that we would run into problems during the implementation process. However, Gemma was better than we expected, and was able to complete the task of designing buildings and finding their cost quite well.”
HomeCrush helps home seekers discover what “home” means to them by designing and exploring a furnished, two story 3D house through conversation with Gemma. Refine colors, flooring, lighting, and décor, then explore real U.S. property listings through a Tinder-inspired experience with photos, hearts, and saved favorites. Traditional searches start with filters; HomeCrush starts with imagination, helping people find what they truly want in a home.
“A simple request like “walnut floors and more plants” could produce a good design idea, but Gemma would have a hard time implementing it in the 3d model since it is a fairly small model The biggest improvement came from adding a validation step to the process so Gemma could review what actually goes in the 3d model, so all the final changes are refined.”
Virginia renters can lose their deposit by missing a 5-day deadline to dispute the landlord’s move-in report. FinePrint turns their lease and phone camera into sealed evidence and deadline reminders to help avoid this. Got a question about your lease? Gemma 4 flags illegal lease clauses and answers questions with cited Virginia law.
“Choosing an iOS app made development extremely slow and difficult. We aspired to include every possible tool that could help a tenant, but with only 8gb of ram on an old mac mini, testing was a nightmare, with some features taking up to 4 hours to ship.”
Disclosure is an open-source website that analyzes home damage and aims to inform the user of how and when to make repairs. This website is for any homeowner that has home damage of all severities. Also, the cool part is that it uses the AI model Gemma 4 along with TypeSafe AI to do the recognizing and categorizing of home damage through user uploaded images. Finally, the main Google tool we used throughout this project was Cloud Run to host the website, Gemma AI, and TypeSafe AI.
“We were pleasantly surprised by how well Gemma 4’s image classification worked with TypeSafe AI to get a consistent level of accuracy.”
FixTrail turns a photo of a housing problem into a dated record, a response clock, and a repair notice, and shows when many households in one building report the same problem. When a repair request goes nowhere, renters often have no proof they ever asked. Complaints are usually made by phone, in passing, or in texts that get lost, so a landlord can say they never heard about it. If five households in a building have the same mold problem, nobody sees that it is one building-wide issue.
“I was surprised at Codex's programming capabilities after using Claude for most of my project development thus far. Codex apparently can now take control of your computer to interact with the browser and able to autonomously navigate it”
Footing is a free app/website for people experiencing homelessness in Virginia, including immigrants whose papers decide which help they can get. You tell it your status, who’s with you, and what matters most, and it ranks all 133 Virginia cities and counties by where you’re most likely to get back on your feet. It shows which programs you can actually use there and where every number comes from. Your answers never leave your phone, there’s no account, and it works offline in 9 languages.
“We understand now why git is a industry standard as we tried to move files using google drive making collaboration 10 times harder. It went so much smoother after using githubs qualities.”
NestWise helps first-time renters understand the true cost of housing before signing a lease. It combines real housing data, financial analysis, and Google Gemini to uncover hidden fees, explain lease terms, and compare rental options. Its privacy first approach lets users redact personal information before AI processes their documents, helping them make more informed rental decisions.
“What surprised me most was how much work goes into data engineering before you can actually build useful features. Working with real housing data taught me how important it is to clean, organize, and connect different datasets, even when they come from different sources.”
Lease Red-Flag Checker helps students, and anyone else renting a place, check the quality of their lease. The website lets users upload a PDF or text, which the program analyzes and turns into highlighted lines of flagged risk ranging from low to high. When users hover over a highlight, they see suggestions and the relevant law, and the site also drafts an email to their landlord. With links to the Virginia housing codes and student legal services, users can also look directly at their rights
“One of the interesting things I had to deal with was that even though the pasted text or uploaded PDF had the exact wording of the lease, the Gemma model would catch the questionable clauses but sometimes reword them when quoting them back instead of using the exact words. That meant the highlights couldn’t find the text on the page, so I had to add fuzzy matching to locate the original clause.”
RoomRoot helps people find housing and housemates who fit their routines, needs, and expectations. Users can build a housing profile or describe a space to share, compare lifestyle preferences, and use conversation prompts to talk through the details. The site also showcases room listings with practical details, an anonymous community for shared-housing advice, and RootSpace, a 3D room viewer for imagining and planning a space.
“I was surprised by how much finding a good home depends on everyday details people don’t usually discuss: quiet hours, guests, cleaning, and shared spaces. RoomRoot made me realize that helping people talk about those expectations early can matter just as much as finding an available room.”
Housing AR allows users to not only find housing listings for them but also experience an immersive 3d experience of setting up their dream home remotely. It is very user-friendly so even people with no experiences in 3D development can easily set up their new home using their old furniture.
“I assumed the implementation of segmentation and reconstruction would be as easy as it was on 2D. It took a painstaking long time of troubleshooting and testing different apps for it to finally be able to reconstruct 3D objects.”
Clear House helps low-income people in Northern Virginia and Blacksburg find verified affordable housing, shelters and rent help, especially students and families in a crisis. Its AI chat answers only from checked resources, and it sends anyone in danger straight to 911, 211 and hotlines. The cool part: ask in plain words and get real phone numbers, walking times and a map of homes you can actually reach.
“We expected the AI to be the hard part, but the real problem was that a housing listing can’t tell you if the place is still open. So we built the whole site around that: every listing has a “Call to Confirm Availability” style button.”
True Cost helps students at VT who are looking for off-campus housing see the actual price of housing after all the fees and be able to compare apartments side by side. Our database of over 200 apartments around VT gives VT Students plenty of options, and they can sort the database to their liking. Our Chrome extension allows them to, instead, choose an apartment that they like and scan the listing page and to find all the fees included and compare it to other apartments in our database.
“Something that surprised me was actually how simple it was to make a google extension. I have never done this before and I thought it was a long and tedious process, turns out it wasn't.”
This project to help people who are getting ready to move. The goal is to help find and budget for a new home! It gathers the cost information (i.e. moving cost, upfront costs, etc...) and (given a timeline of when you need to move) will create a budget to help you save up enough money to move!
“At first I thought that I would just have to hook up Gemma and it would do all the calculations and searching for me. Now I know that Gemma just really handles the input and what to do with it but the heavy calculations and searches are built separately and Gemma calls those.”
True Cost of Rent reads a rental listing from a screenshot or link and works out what the place really costs each month, including the utilities, fees and commute the listing leaves out. It flags every charge the listing never mentioned and writes the questions to ask before you sign. It is built for first time renters, and across ten real listings the advertised price understated the true cost by 23 percent. Gemini reads and explains, and Python does all the math.
“I wrote a rule telling Gemini not to touch the roommate count on shared listings, and it changed it anyway, which doubled a rent share and made one place look twice as expensive. I found it by testing my own app, and the fix was moving the rule out of the prompt and into code.”
HomeCrew lets renters video-call an AI crew and just point their phone at what’s broken. Rosa spots the problem, Theo tells you which Virginia law makes it your landlord’s job, and June writes the repair letter for you. It’s for Virginia renters, who almost never have a lawyer when something goes wrong. It also tells you when something is yours to fix and how, or when you need a pro.
“Gemma kept calling the hole in our friend’s door a bug on the wall. Getting closer made it worse; what fixed it was filming from farther back with the doorknob in frame, so context mattered more than detail.”
Submissions for this jam closed at the end of Sunday, October 25, 2026.
