~/index · /about

Builder. Author.
Reluctant futurist.

Engineer at the seam where language models meet production systems. I write code, I write books, and I run my mouth on stages — usually about why agents need ops too.

I build AI agents that ship. Not the demo-reel kind — the kind that operate inside enterprise systems with real users, real audit logs, and real stakes.

Currently GenAI Engineer at Zato, where I lead agent architecture across our document intelligence (DocIQ) and operations copilot (OMS) products. Before that, I cut my teeth on data science, computer vision on drones, and Web3 health records.

I'm also an authorFrom Data to World is my attempt to write a data-science book that doesn't read like a tutorial. I speak at meetups (Microsoft Reactor, ThoughtWorks, Solapur FOSS, TinyML), mentor through Microsoft's MLSA program, and run education projects at Atria Foundation.

Outside of work I read systems papers and rebuild old ideas in modern stacks — to see what changes when the building blocks get an order of magnitude better.

Currently
GenAI Engineer @ Zato
Based
Bangalore, Karnataka, India
Studying
B.Tech, Digital Transformation · Atria University · 8 CGPA
Writing
From Data to World · blogs.lakshaykumar.tech
Speaking
Microsoft Reactor · ThoughtWorks · Solapur FOSS · TinyML · 15+ talks
Mentoring
Microsoft Learn Student Ambassador · Atria Foundation
Reach
[email protected]
[01] · stack

What I work with.

/agents

Agent frameworks

LangGraph for stateful multi-agent flows. Anthropic Claude SDK for production reliability. Custom orchestrators when frameworks get in the way.

langgraphanthropicopenaillamaindex
runtime
/data

Retrieval & data

Hybrid search, re-rankers, structured extraction. Postgres pgvector for low-ops, Qdrant/Pinecone when scale demands it.

qdrantpineconepgvectorelasticcohere-rerank
storage
/serve

Serving & infra

FastAPI + Pydantic for typed tool surfaces. Docker, GCP/Azure, Redis for state, Postgres for source-of-truth. Streaming everywhere.

fastapidockerpostgresredisgcpazure
production
/web

Web & experience

React/Next for agent UIs — streaming chat, tool-call visualizations, audit trails. Tailwind for speed when shipping is the brief.

next.jsreacttypescripttailwind
frontend
/foundations

Classical foundations

Where I started: scikit-learn, classical ML, computer vision (OpenCV, MediaPipe), Power BI dashboards, statistics. Still pull these out when LLMs are the wrong tool.

scikit-learnopencvmediapipepandasstatistics
classical
[02] · receipts

Education & certifications.

2021—

B.Tech · Digital Transformation

Atria University, Bangalore — current cohort. 8 CGPA. Cross-disciplinary curriculum spanning ML, design, and systems thinking.

undergraduate
2020—21

Class XII · ISC Non-Med

Jiva Public School — 91%.

secondary
/cert

Certifications

Python for Data Science · IBM. Data Analysis with Python · Cognitive Classes. Microsoft AI Classroom Series.

ibmcognitive-classesmicrosoft
credentials

Want to compare notes?

Tech discussions, mentorship, or speaking — happy to chat. I reply to every thoughtful note.