TATO | Devpost

Architecture diagram

Inspiration

TATO came from a job I worked where I hauled items away for people and realized how much value was being thrown out every day. A lot of what we picked up had been acquired for free after hauling jobs, and when I started reselling those items, it opened my eyes to how big the resale market really is.

What stayed with me even more was how customers reacted. A lot of them appreciated that their old items were finding a new home instead of ending up in a landfill. That made me see recommerce as more than just flipping inventory for profit. It also has real environmental value. TATO was inspired by that experience and built to help usable secondhand goods stay in circulation longer.

What it does

TATO is a recommerce platform that helps suppliers intake secondhand inventory faster and helps brokers discover inventory that is ready to move.

For this challenge, the heart of the product is a live supplier intake agent powered by the Gemini Live API. A supplier can point their camera at an item and talk naturally while the agent watches, listens, responds in real time, asks follow-up questions, and can be interrupted just like a real conversation. While the conversation is happening, the agent continuously builds a structured draft with the item title, brand, condition, missing views, floor price, and suggested listing details.

Once an item is ready, it can move into TATO's broker workflow, where brokers can evaluate opportunities, claim inventory, and continue through listing and settlement flows.

How we built it

We built TATO as a cross-platform app using TypeScript, React Native, Expo, and Expo Router so it could run across mobile and web. The backend uses Supabase for authentication, database storage, row-level security, and serverless functions.

To support the live agent experience, we built a Google Cloud-hosted live agent service that uses the Google GenAI SDK to create short-lived Gemini Live API sessions securely. That lets the client connect to a real-time voice-and-vision agent without exposing long-lived API keys on device. The live agent is also tool-enabled, so it can update the UI with structured intake state while continuing a natural spoken conversation.

We also integrated Stripe and Stripe Connect for claim fees, payments, and payout workflows, and used Sentry and PostHog for monitoring and analytics.

Challenges we ran into

One of the biggest challenges was making the live experience feel natural while still keeping the output structured and useful. We did not just need a chatbot. We needed a real-time agent that could understand voice and camera input, handle interruptions, ask the right next question, and still produce marketplace-ready data.

Another challenge was security and reliability. Because this is a live multimodal experience, we had to think carefully about authentication, short-lived session bootstrapping, browser and device permissions, fallback flows when live intake is unavailable, and how to keep latency low enough for the interaction to feel helpful.

We also had to design for two different users: suppliers, who need fast and simple intake, and brokers, who need enough clarity and confidence to act on an opportunity.

Accomplishments that we're proud of

We're proud that TATO is rooted in a real-world problem and not just a hypothetical use case. It came directly from hands-on experience in hauling, resale, and seeing how much usable inventory gets overlooked.

We're also proud that we built a live multimodal workflow around a real business problem. Instead of using AI as a novelty, TATO uses the Gemini Live API to make secondhand intake faster, more conversational, and more actionable. The result is a system that connects live voice-and-vision intake, structured item records, broker discovery, and settlement flows in one product.

And beyond the tech, we're proud that the project supports a more sustainable outcome by helping items get reused instead of discarded.

What we learned

We learned that the resale problem is not just about finding buyers. A huge part of the challenge is capturing inventory quickly and clearly enough for the next person in the chain to trust the information and take action.

We also learned that live agents are most powerful when they are grounded in a real workflow. The Gemini Live API becomes much more useful when it is not just talking, but actively helping move a real task forward through voice, vision, tool use, and structured outputs.

What's next for TATO

Next, we want to make TATO even stronger at live intake quality, pricing intelligence, crosslisting, and listing generation so suppliers can go from discovery to broker-ready inventory even faster.

We also want to expand the live agent experience with better multilingual support, smarter follow-up prompts, and deeper broker-side intelligence around demand and matching. A big part of the roadmap is improving crosslisting so inventory captured once in TATO can be prepared and pushed more efficiently across multiple resale channels.

Long term, the goal is to make TATO a full recommerce operating system that helps people recover value from overlooked inventory while reducing waste.