SilentPartner

Inspiration

We've all been there. You drop $1,500 on a ticket to a massive conference like TechCrunch Disrupt. You walk into a room of 10,000 people with a clear goal: find a technical co-founder, meet an investor, or just have some genuinely good conversations. But the reality? You grab a coffee, awkwardly wander the expo floor, try to read tiny nametags without staring, and just hope you get lucky. It’s overwhelming, inefficient, and honestly, pretty anxiety-inducing. We realized that at these events, you are surrounded by incredible opportunity, but the friction of figuring out who to talk to and how to break the ice stops you in your tracks.

What it does

That’s exactly why we built SilentPartner. It’s a React Native mobile app that acts as your real-time, social coach right in your pocket. Instead of leaving your networking up to chance, SilentPartner cuts through the chaos and curates a personalized list of high-value connections. We don’t just give you a directory of who is in the building; we pinpoint exactly who you need to meet, and we act as your wingman giving you the real-time coaching and exact scripts to start the conversation.

How we built it

To bring SilentPartner to life, we built a fast, cross-platform mobile app using React Native and Expo. Our backend runs entirely on Firebase, which securely handles our database and the seamless LinkedIn onboarding. The real magic happens in our audio stack we use WebRTC to stream live voice data directly into the OpenAI Realtime API, which acts as the brain giving you instant, conversational coaching. Finally, to make sure the app doesn't drop the ball in a crowded conference hall, we monitor all our real-time performance and reliability using Embrace.io.

Challenges we ran into

Building a real-time AI app in just one weekend meant our biggest overarching challenge was simply time management and knowing what features to prioritize. Our first major technical hurdle was the LinkedIn integration, as navigating their strict OAuth flow to cleanly extract user data took much longer than we planned. However, the absolute hardest roadblock was getting the live audio to work properly without crashing the app. We spent hours fighting with WebRTC and Expo to establish a stable, low-latency audio stream, but we eventually got it working smoothly with the OpenAI Realtime API.

Accomplishments that we're proud of

We are incredibly proud of building a fully functional, real-time AI voice app in just a single weekend. Stitching together WebRTC, Expo Audio, and the OpenAI Realtime API into a smooth, crash-free experience was a massive technical win for the team. Beyond the code, we’re really proud to have built an intuitive, fast-feeling UI that genuinely solves a real human problem the overwhelming anxiety of networking at massive events.

What we learned

Technically, we learned a ton about managing low-latency audio streams and handling complex OAuth flows in React Native. On a personal level, I've always struggled with an inability to decline requests, which often leads to taking on way too much work. This time-crunched weekend forced a change in habits. Learning to say no to "nice-to-have" features so we could focus entirely on delivering a polished, working core product was our biggest takeaway.

What's next for SilentPartner

The immediate next step is beta testing the app at a live, local meetup to gather real user feedback on the AI's real-time coaching. Looking further ahead, we want to expand beyond public conferences and explore enterprise use cases potentially integrating with corporate architecture tools like SAP LeanIX to help map and facilitate internal networking for massive, siloed organizations.