Who I am

I'm Heet — an AI engineer and builder based in India, working at the intersection of machine learning, developer tooling, and systems design, with an eye toward privacy and clarity.

I'm a student, but I spend my time the way an engineer would: shipping projects, contributing upstream, and writing down what I learn so it's useful later.

The best architectures come from constraints, not freedom. A single-machine stack, a privacy-first requirement, a one-week hackathon deadline — these aren't limitations, they're guardrails that keep the design honest.

I'd rather have a few abstractions that compose cleanly than many features that never quite touch ground. Good systems survive their authors leaving; most don't, and the difference is deliberate simplicity.

Before optimizing a model, I trace where the latency and complexity actually live — often it's not the slowest-looking part, but the handoff between two systems that neither one owns.

I write for the future reader, including future me. Code and notes are the same thing: a message to someone who's forgotten the context.

I build because the things I want usually don't exist yet — and the things that do exist rarely fit how I want to work.

The most satisfying work sits quietly between a few people for a long time. A tool that saves a team a minute a day compounds into weeks saved.

Rust is teaching me to treat memory and performance as design constraints, not afterthoughts — useful for the audio path in Tatvik.

The MCP ecosystem is moving fast. I'm reading the spec, building tooling against it, and revising my assumptions with every release.

I'm deepening work on local-first AI: agents that reason about your own tools without leaving your machine, plus the documentation that helps others adopt that model.

Everything else feeds one question: how do we make intelligent systems people can trust by default?

I started in electronics and IoT, which gradually became a deeper interest in AI and machine learning. My path has moved roughly: IoT → Web Development → Full-Stack Engineering → AI Applications → Machine Learning → NLP & LLMs → AI Systems → Open Source.

I like understanding systems end-to-end rather than working on a single layer — how data is collected and stored, how a model predicts, how it's wired into an application, how it scales, and how people actually use it.

My current foundation is machine learning, with NLP as my chosen domain: supervised and unsupervised learning, neural networks, TensorFlow, Scikit-learn, plus large language models, retrieval and QA systems, AI agents, LLM pipelines, model inference, and AI-powered developer tools.

I care about moving past just consuming AI APIs, toward understanding the full pipeline: Data → Preprocessing → Model → Training → Evaluation → Inference → Application → Deployment.

Machine Learning — linear algebra, probability, statistics, optimization, classical algorithms, model evaluation, feature engineering.

Deep Learning — neural networks, transformers, transfer learning, representation learning.

NLP & LLMs — transformers, tokenization, embeddings, retrieval, QA, RAG, agents, inference systems.

ML Engineering — data pipelines, experiment tracking, model serving, deployment, monitoring, optimization.

AI/ML Engineering, NLP, LLM Systems, AI Agents, Developer Tools, Open Source, AI Infrastructure, Intelligent IoT, Real-Time AI Applications.