Past Clients

What we shipped

The pod rebuilt RePackify's notification layer so shipping and order alerts reach users on web and mobile alike. They then extended the app's AI assistant: it now handles multi-turn search, creates product listings, and can send someone straight to checkout by voice or text. Both sit in one responsive interface, backed by pipelines that keep inventory and transaction state current.

What we shipped

The pod built a ticketing system that runs itself off GitHub. Incoming issues get scored on effort against priority, ranked, and assigned a micro-bounty, with fallback rules for when the score comes back inconclusive. Webhooks and REST calls keep a local mirror of repo state, and an LLM pass rewrites thin or unprioritized tickets into something a developer can act on.

What we shipped

REFIND needed two front ends, so the pod built both: a workspace where brands run things, and a mobile-first view for their customers. A brand can spin up an event, assemble a custom form for it, watch the numbers come in live, and run the loyalty program the team named giftpath. An LLM layer drafts those feedback forms and summarizes how an event went.

What we shipped

The hard part here was getting the device and the phone to talk. The pod wrote the Bluetooth Low Energy link between them, along with firmware controls that stream sensor readings live, buffering so a dropped packet doesn't lose data and caching locally when the connection goes. On the app side they built the charts that draw the incoming bio-signals and the sync path back to the server.

What we shipped

Matching a customer to a chef was being done by hand, so the pod automated it. A request first passes a filter for distance and dietary coverage; whatever survives gets ranked by cosine similarity between embeddings of the request and each chef's profile. The customer sees the top five.

What we shipped

The app's data was scattered across text files and CSVs, so the first job was moving all of it into Supabase. From there the pod wrote a weekly job that pulls events from APIs and scripts, cleaning up duplicates and formatting on the way in. Recommendations run on a RAG setup: embeddings written on insert by trigger, then vector search through Postgres RPC calls.

What we shipped

The pod built the capture side on a Raspberry Pi: buffer the mic input, split the stream, push it to a server for transcription and keyword detection. They then wrote the models that check an impression against the video and tie it back to the transaction, which is what turns a passing mention into a number anyone can bill against. The backend runs in containers on Google Cloud Run.

What we shipped

The pod shipped an MVP to the App Store, with onboarding, org profiles, and event notifications behind it. Most of the semester went into the recommender that matches a student to orgs based on their major, their hobbies, and what they want to do after graduation. The point was to replace the group chats and flyers people currently rely on to find any of this.

What we shipped

Across two semesters the pod built the models that place furniture into an empty room, pairing convolutional networks with heuristics for where a piece should sit. A second model runs the other direction: a stable diffusion inpainting pass that strips the furniture out of a photo of a furnished room. Both are served from Google Cloud endpoints.

What we shipped

The pod built the help section of the First Principles app: FAQ panels, plus an AI chatbot for the questions the FAQ doesn't cover. They also mapped the user flows for the rest of the app and designed more than fifteen screens across low and high fidelity.

What we shipped

Convergent has worked with Jump Finance for four semesters now. This one went to course registration: students can sign up for classes inside the Jump Finance portal instead of leaving it partway through. The pod also expanded the portal's resource section so academic and financial support material is easier to find.

What we shipped

The pod built a cross-platform desktop toolbar that gathers a handful of small utilities in one strip: transparent sticky notes, screen warmth that shifts through the day, backgrounds that swap themselves before a call, and a warning when you start talking while muted. Settings survive a restart, and a Windows installer handles setup in a few clicks.