Production agentic AI
Architected and built multi-agent systems for SRE investigations, bringing together dynamic specialist routing, MCP integrations, service topology, and asynchronous analysis.
SOFTWARE ENGINEER
Madhulika
Maddipudi.
Building intelligent systems.
Engineering them to be reliable.
I work at the intersection of AI, distributed systems, and observability—turning complex technical problems into software engineers can depend on.
Explore my work01 — ABOUTCuriosity in the details.
Curiosity in the details.
Clarity in the system.
I’m a software engineer with over eight years of experience building enterprise software across SaaS, consumer devices, and networking.
My work spans production AI systems, engineering automation, and platform reliability. I build the infrastructure that helps teams develop, evaluate, release, and operate software with confidence.
I’m especially interested in the engineering around AI: how agents choose their tools, how systems behave when dependencies fail, and how we know an investigation actually reached a useful outcome.
8+years connecting deep systems work
with practical engineering outcomes
with practical engineering outcomes
02 — SELECTED ENGINEERING
Complex problems.
Considered solutions.
A selection of the systems I’ve built
and the problems I care about.
Reliability by design
Designed agent execution patterns that separate deterministic control from model reasoning, constrain actions to runtime context, and isolate downstream failures. Built extensible integrations that accommodate new providers without expanding a central agent.
Evaluation beyond the answer
Built AI evaluation infrastructure that examines the full execution trajectory: routing, tool selection, parameters, dependencies, and task completion. Combined deterministic checks with model-based evaluation to assess how agents reach their conclusions.
Engineering at every layer
Delivered software across distributed backends, Linux and embedded systems, and networking. Built automation and release engineering solutions that help teams develop, validate, deploy, and operate enterprise software.
TOOLS OF THE TRADE
Python · Java · TypeScript · MCP & APIs · Linux
03 — WRITING
Notes from
the engineering.
Ideas and lessons from building systems
that need to work in the real world.
What Happens When You Give an AI Agent Access to Several Different Monitoring Tools
On orchestration, evaluation, and the engineering that makes agentic systems dependable.
Read the article on Medium (opens in a new tab)