Anton Medvedev

I helped design and launch a LangGraph-based AI support chat platform for a high-volume fintech product, taking it from early planning to rollout for 100% of users. To ship the MVP quickly, I pushed the team away from premature microservices and toward a simpler first version that could reach production fast. I built the Salesforce handoff flow for transferring chats to human agents, designed the service-to-service and end-to-end APIs, introduced an AsyncAPI spec for front-end and back-end coordination, and made the Socket.IO communication layer scalable.

Later, I designed the target microservice architecture using Kafka to balance chats across pods and isolate stateful logic in a dedicated service, enabling the rest of the system to scale horizontally. I also tuned the RAG pipeline and pgvector performance, improved deployment reliability, and introduced linting and testing standards. The result was a chatbot that answered with accuracy comparable to real support agents, improved CSAT by more than 20% over the previous bot, and helped support scale without linear headcount growth.

I built a Chrome extension that merges PDFs locally in the browser, with privacy, speed, and reliability as the core product goals. I shipped the first version in about two weeks during a micro-products bootcamp, building the front end in TypeScript and React and using AI-assisted tooling to move quickly despite not being a dedicated front-end specialist.

I evaluated PDF libraries and implementation approaches, optimized the merge pipeline, and refined the UX so large file combinations felt almost instantaneous while keeping documents on the user’s device rather than uploading them to the cloud. I also partnered on product design to deliver a clean interface, then used Amplitude analytics and SEO work to improve activation and organic growth after launch. Within a year, the extension grew to nearly 1,000 users, earned featured status, and maintained a perfect rating.

I worked as a full-stack engineer to develop Free2Shred from the ground up, an iOS app for the street sports community. The product combined a community-driven spot map, media features, and social discovery into one mobile experience for skaters. I built the back end API in Python and FastAPI, created the integration layer for the Flutter client, and implemented core mobile features, including screen flows, animations, navigation, and media handling.

I also integrated Google Maps, Firebase, analytics, and other 3rd-party services to support location-based discovery and reliable content workflows. A major part of the work was delivering video editing, upload, and playback so users could create and share content inside the app instead of relying on external tools. In parallel, I set up the development and deployment environment with Docker and CI/CD to keep delivery predictable. The result was a polished community product that brought mapping, content creation, and mobile experience design together in one app.