About
GeekyRahul builds AI agents as personal learning projects and writes about what he learns.
I build AI agents as a way of learning, and I share the code, the costs, and the failures publicly. Everything here is a personal learning project, built on my own time with my own tools and only public or synthetic data. I don't offer services, and nothing on this site is a product or an offer of work.
How I think about agents
An agent is only interesting if it survives contact with real inputs. So every project here ships with the same evidence: an architecture diagram, measured cost and latency, an evaluation result, and an honest list of what broke. Demos are easy; reliability is the actual subject.
Skills and tools
Agent building
- LLM tool use and function calling
- Retrieval-augmented generation
- Evaluation harnesses
- Prompt design
Engineering
- Python
- TypeScript
- FastAPI
- PostgreSQL
- Docker
Practice
- Cost and latency measurement
- Failure analysis
- Building in public
Journey
I started publishing agent experiments in 2025 under the series name "We4AI Lab". Each entry is one agent, built end to end, with the numbers recorded. The goal is a public trail of competence: verifiable, dated, and honest about failures.
How I work
Small scope, measured results, written up quickly. Every project page follows the same template so readers (and machines) always know where the evidence is.
Elsewhere
- GitHub — all project code
- LinkedIn — the same name and bio
- hello@geekyrahul.in
Views are my own. Projects here are personal, non-commercial learning projects.