I work on backend systems where correctness and security aren't an afterthought — mainly within the Frappe ecosystem (ERPNext, HRMS), where I've spent time auditing and hardening production code. My focus areas are permission/authorization logic, data lifecycle correctness, and error handling that fails safely instead of silently.

Alongside that, I build with LLM-based systems — retrieval-augmented generation, agentic workflows with LangChain and LangGraph, and applying the same rigor from systems work to how these pipelines are designed and evaluated.

• Backend & Framework Internals — Python, Frappe, MariaDB
• Security & Correctness — permission checks, lifecycle/state handling, defensive error handling in large codebases
• Applied AI — RAG pipelines, agent orchestration with LangChain and LangGraph
• Open Source — contributing fixes upstream rather than working around issues locally

I read code like both an attacker and a maintainer — every change is grounded in the system's actual contract and behavior, not assumptions. I'd rather ship a small, well-reasoned fix with a clear reproduction path than a large, unreviewable one.

Building systems that hold up under scrutiny.