_{repositories  ·  hugging face  ·  live demo — ai companion  ·  live demo — web app  ·  portfolio hub  ·  contact}

Independent by habit, engineer by discipline.

I'm Rishi Raj DEV — I work where AI research meets production software. Calling a model API is the easy part, so I spend my time on what sits underneath it: training loops, latent representations, evaluation, and the plumbing that turns an experiment into something people can actually use.

On the application side that means realtime interfaces, typed APIs, databases that stay honest, CI that runs, and deployments I can point at. I prefer systems I can explain end to end and claims I can back with evidence.

Quick facts

• Focus — Generative AI · diffusion models · full-stack engineering
• Daily drivers — Python · PyTorch · TypeScript · React · Node · PostgreSQL
• Currently — Research engineering and low-compute model training
• Collaboration — With @Mrquillwing on Ojasvat
• Based in — India · UTC+5:30 · working worldwide
• Open to — AI/ML roles, internships, research collaboration
• Reach me — rishirajc406@gmail.com

Note

I would rather show you the failure than oversell the result. Most of what I build is measured against a baseline I can defend, and written up whether or not it worked. A model that produces beautiful colour but broken geometry gets described as exactly that — not as a breakthrough. If we work together, you will always know which state a thing is in.

_{AI / ML — modelling · training loops · evaluation · notebooks · artifacts}

_{WEB / REALTIME — typed frontends · APIs · WebSockets · 3D in the browser}

_{DATA / PLATFORM — schemas · CI · containers · deployment · reproducibility}

_{one star per night I shipped · size = volume · lines = consecutive nights · this sky is my actual commit history, so it is one-of-one by construction}

_{Activity cards come from third-party services and can be briefly unavailable, delayed or rate-limited. Counts are dynamic — GitHub itself is always the source of truth.}

• 🔬 Research — Generative models, training stability, latent representations, evaluation, low-compute experiment design — and honest write-ups of what did not work.
• 🛠️ Build — AI-powered products, realtime features, typed APIs, 3D interfaces, data pipelines and the deployment plumbing that turns a demo into a product.
• 🤝 Early-career — Internships and junior AI/engineering roles where I can contribute hands-on from week one, review code seriously and keep improving in public.

Build carefully. Measure honestly. Keep shipping.