Sagar Kandpal — GenAI & full-stack developer. I build agents, RAG systems and the products around them.
Every number here comes from the project itself. No vanity metrics, nothing invented for a portfolio.
I’ve never believed in waiting until I have everything figured out. I’d rather start, experiment, break things and learn on the way — agents, RAG systems, full products, even the videos about them. Progress comes from doing, not just knowing.
Seven AI products, built end to end. Each one told properly: the problem, what I built, and the part that was actually interesting.
A FastAPI agent that transcribes, translates, dubs and summarises any video.
Chat with any video — transcribed, embedded and searchable. Scroll through the pipeline.
Paste a video link. The assistant pulls the source.
Strip the audio track — that’s all the model needs.
Whisper turns speech into timestamped text.
Split the transcript into overlapping chunks that keep context intact.
Every chunk becomes a vector.
Vectors go into FAISS / ChromaDB for fast similarity search.
Ask anything. Retrieval finds the right moments; the LLM answers from them.
Deep on retrieval, comfortable across the full stack. Tap any tool to see where I’ve actually used it.
Most homes are full of appliances that aren’t smart — and replacing them all with "smart" ones is expensive. So my friend Arjun Purwar and I asked a simpler question at a hackathon: why not make the ones you already have smart?
We built it around an ESP32 and a 4-channel relay. You can switch appliances from your phone over Wi-Fi, a PIR sensor turns the light on when someone walks in, and a push button still works when there’s no network — that’s the "hybrid" part.
ESP32 · 4-channel relay · Wi-Fi / MQTT · PIR sensor · Push button
