I've built full-stack AI and 3D products at Inbenne, including a Docker Compose-based seven-container platform integrating FastAPI, MongoDB, Redis, Celery, and Ollama.
I improved real-time terrain rendering by moving CPU-bound operations into custom GLSL shaders, taking frame times from about 83 ms to 16 ms for stable 60 FPS. I also designed a Celery and Redis task queue that reduced ML-pipeline API acknowledgment from roughly 120 seconds to under one second.
At IIT Roorkee, I developed Linear and Logistic Regression, Decision Trees, and k-NN from scratch, matching scikit-learn predictions within 0.1% mean absolute percentage error. I also built preprocessing pipelines that cut data-preparation time by 60% and contributed 10+ merged pull requests through GirlScript Summer of Code.
My projects include OVO, a Git-like music version-control system, and Virtuoso Vision, an AI-powered 3D audio visualizer. I enjoy combining FastAPI, Next.js, graphics tooling, ML models, and data systems to create interactive products that perform reliably in real time.
