Software engineer from Nairobi, Kenya specializing in retrieval-augmented generation and full-stack products. I take AI systems end-to-end — document ingestion, chunking, vector embeddings, FAISS semantic search, cited LLM answers — and the React frontends that make them usable. My flagship RAG assistant runs entirely on CPU with 100% source citation. Final-year B.Tech IT student at the Technical University of Kenya.
- 🔭 Currently building: a 10-capability local RAG platform — hybrid search, knowledge graphs, corrective retrieval, multi-tool agents
- 💼 Shipped for real businesses: CRM/ERP rewrite at Aquashine (450 files → Flask, 99.9% uptime), automation for a furniture platform and a retail shop
- 🛰️ Also building: Lema-OS, an x86_64 hobby operating system in C
- 🎯 Open to: junior software engineering / AI roles · contract RAG work · open source collabs
- ⚡ Fun fact: I mentor Maasai youth aged 13–17 in coding — and I build WiFi billing systems for fun
| ⭐ | Project | What it does |
|---|---|---|
| ⭐ | prompt-toolkit | Dataclass-based prompt template toolkit — validated placeholders, 3 real RAG prompts (rewrite → cited answer → grade). Week 1 of my AI engineering course. |
| 22 | tuk-convosearch |
Production-grade RAG student assistant — ingestion → chunking → 384-dim embeddings → FAISS → cited answers, fully on CPU. Final-year project. |
| 1 | metadata-rag |
Fully local 10-capability RAG platform — hybrid search, knowledge graphs, corrective retrieval, multi-tool agents. Nine ADRs, golden Q/A suite. |
| 1 | lema_wifi |
Self-hosted WiFi billing system — vouchers, payments, hotspot management. |
| 1 | lema-os |
x86_64 hobby OS from scratch in C — bootloaders, memory management, the works. |
| 1 | markdown-chunker |
Token-aware Markdown chunker built for RAG pipelines. |
AI/ML: RAG · LangChain · LlamaIndex · FAISS · Sentence Transformers · Ollama · NLP
The fastest way to reach me is email — I reply within 24–48 hours.
Always building. Always learning.