I'm an AI Engineer who owns the whole lifecycle: data → model → API → deploy → monitor. Most "ML projects" die in a notebook. Mine ship, get containerized, get observability, and keep working at 3 a.m.
- 🧠 Deep learning & LLMs — I've built a GPT-style transformer from scratch (tokenizer → attention → training loop → sampling), so I know what happens under the API call.
- 🤖 Agentic systems — supervisor / specialist multi-agent architectures with LangGraph, tool use, and human-in-the-loop workflows.
- ⚙️ Production engineering — FastAPI services, PostgreSQL, Docker, Kubernetes, AWS, and Terraform. Typed, tested, observable.
- 🔬 Currently cooking — an LLMOps platform with auth, migrations and eval loops, a local-first knowledge-graph explorer for the history of ideas, and a simulated community of AI-research agents that argue their way to a spec.
| Project | What it is | Stack |
|---|---|---|
| llm-from-scratch | Full GPT implementation in pure PyTorch: BPE tokenization, causal multi-head attention, transformer blocks, training pipeline, greedy / temperature / top-k generation. | PyTorch Python 3.13 uv |
| prefitor | LangGraph multi-agent fitness advisor. A supervisor orchestrates training, nutrition and cooking specialists into one personalized plan. | LangGraph LangChain Streamlit |
| information-retrieval | Academic abstract fetcher that unifies 9 scholarly APIs (PubMed, OpenAlex, arXiv, Semantic Scholar…) behind one interface. Corpus building for RAG. | Python REST APIs |
| mutant_detector | Dockerized REST API that classifies DNA sequences as mutant / human, with stats endpoint. Magneto approved. | Python Docker REST |
| Clinical-trials-on-cancer-data | NLP on patient illness records to predict potential cancer risk. | NLP Scikit-learn Pandas |
| Employee-desertion-detector | Predicting employee churn from HR data, end to end. | Pandas Scikit-learn |
AI / ML
Backend & Data
Infra & Ops
| Repo | What | Last push |
|---|---|---|
| llm-fine-tuning | Fine-tuning experiments across different open LLMs | 8 days ago |
Employee-desertion-detector Python |
Predicting employee churn from HR data with scikit-learn, end to end | 8 days ago |
Clinical-trials-on-cancer-data Python |
NLP on patient illness records to predict potential cancer risk | 8 days ago |
mutant_detector Python |
Dockerized REST API that classifies DNA sequences as mutant or human, with a stats endpoint | 8 days ago |
information-retrieval Python |
Academic abstract fetcher unifying 9 scholarly APIs (PubMed, OpenAlex, arXiv, Semantic Scholar…) behind one Python interface | 8 days ago |
Open to conversations about LLM systems, agent architectures, and getting ML into production for real.
