About Me

I'm acting AI architect for enterprise solutions, with 18 years in software. My work is agentic AI in a regulated payments environment: getting engineering teams to use AI agents without giving up human accountability.

What I Do

I take a problem from "teams want agents to do this" to an architecture, a working implementation, and the standards other teams build on. I write the code myself, mostly in Python and TypeScript. Much of the work is authentication, audit trails, test coverage, and deciding what happens when an agent is wrong.

Platforms I created and architected:

• A governed agent delivery pipeline. It takes a change from intake to a reviewed merge request, with human decision gates along the way. Agents run in isolated workspaces with the requester's own permissions. Approvals are immutable, and the design forbids auto-merge and auto-deploy.
• An agent-agnostic AI SDLC framework. It gives Claude Code, Codex, and Copilot the same governed onboarding, project knowledge, skills, and gates, so results stop depending on which team or repository you happen to be in.
• An enterprise MCP SDK. One secure-by-default server shape: mandatory JWT authentication, origin validation, rate limiting, SSRF-safe outbound allowlists, and a compliance checker that runs as a required release gate.
• An evaluation platform for skills, agents, and MCP servers. Fast static gates run offline with no keys and no network. LLM-based evaluation runs separately. Releases are gated on contracts.

Around these I've built agent memory, an event-driven bridge from tickets to agent sessions, an AI change-request evaluator, and a predictive scoring pipeline with XGBoost on Snowflake.

Open Source

I build public tools around the same problems I write about. One of them is subagent-evals, an open-source quality index for markdown-defined AI agents. It lint-checks, scores, and replays agent instructions across Claude Code, Codex, GitHub Copilot, Cursor, and Windsurf, with CI-style feedback for teams treating agents as maintained engineering assets.

Teaching

I founded and run an internal AI engineering academy: masterclasses on coding agents, RAG, and spec-driven development, plus regular sessions and workshops. The public side of that work is the tutorials on this site, including tracks on LangGraph, LLM systems, and RAG from first principles to release gates.

Background

Before AI platforms I spent years in the systems they now have to fit into. At Fiserv I led cloud migrations and built microservices for payment applications. At CGI I worked on train planning systems for Deutsche Bahn. At Capgemini I led a SEPA compliance programme across 22 applications.

I hold a Post Graduate Diploma in Machine Learning and Artificial Intelligence from NIT Warangal and a Master of Computer Applications, along with TOGAF, AWS Solutions Architect, and Docker certifications.

This Website

I write about agent architecture, evals, and running AI in production. If you're working on similar problems, get in touch.