Back to work

Case study

SynapseAI

Full-stack AI assistant for chat, PDF analysis, and image/video generation, with streaming responses and persistent sessions.

JavaScriptNode.jsExpressGroq APIGenerative AI

Problem

Wanted a single AI assistant that could hold streaming conversations, reason over uploaded documents, and generate images and video — without being locked into one model provider.

Approach

An Express backend proxies Groq's Llama 3.3/3.1 models (with a Hugging Face inference fallback) behind a unified chat API, supports server-sent streaming and in-memory session history, and parses uploaded PDFs (multer + pdf-parse) for document Q&A. The frontend is vanilla JS with markdown rendering and syntax highlighting.

Code