About
Bhaskar Agisetti: Systems Architect, Builder, and lifelong learner.
I'm Bhaskar Agisetti. I architect systems, build them, and figure out how to make them actually useful to the people who depend on them. My day-to-day spans AI development, data engineering, enterprise integration, and backend systems. The common thread across all of it is curiosity about how things work at a fundamental level.
What I've been building lately
Most of my recent work is in Python, FastAPI, and FastMCP. I've been building semantic search systems that compare pre-indexed search against live search, then evaluate the results using LLM-as-a-judge. That last part is interesting: instead of writing deterministic evaluation logic, you point an LLM at both result sets and ask it to reason about which one is better and why. It gives you a qualitative signal that's hard to capture with metrics alone.
On the infrastructure side, I'm exposing tools as MCP servers, building telemetry into those systems, and surfacing metrics to stakeholders via Databricks. The data engineering side of that involves designing medallion pipeline architectures and deploying across workspaces using Databricks Asset Bundles. One thing I built that stakeholders actually use: Genie dashboards that let them run natural language queries against extracted data without needing SQL.
A lot of what I do involves getting data out of silos. Collecting structured and unstructured data from different enterprise systems, bringing it into a unified integration layer, transforming it into something meaningful, and building dashboards that give people actual visibility. Recently I went deeper into iPaaS (Integration Platform as a Service), specifically MuleSoft and the AnyPoint Platform. The question that keeps pulling me in is: how do enterprise systems actually connect to each other at scale, and what does a connector really do under the hood?
I've also built a lot of API integrations across enterprise systems, and I have deep domain knowledge in healthcare: eligibility verification (270/271 EDI), prior authorization workflows, and EOB processing. That came from genuine curiosity about the domain rather than it being required for the work. Domain knowledge is not a blocker for delivering meaningful solutions, but having it changes the quality of what you build.
How I work
I want to understand the primitives, not just the abstractions. When something breaks or behaves unexpectedly, I want to trace it all the way down. Why does switching a docker context change which daemon your CLI talks to? What is actually happening when a VM spins up in under a second? These questions are more interesting to me than memorizing API surface areas.
I also self-host wherever it makes sense. I run a Mac Mini as a personal server. LawnServe, an operations platform I built for a lawn care business (customer management, scheduling, invoicing), runs on it. It keeps me sharp on end-to-end delivery without managed services hiding the complexity.
Why this blog exists
This is my working wiki. When I go deep on something, I write it down here. If I can't explain it clearly, I don't really understand it yet, so writing is part of the process.
The longer-term idea: vectorize these notes and build a personal bot that can answer questions from them. A conversational interface into my own knowledge base. This blog is the data source for that.
If someone else finds the articles useful, good. But I write them for myself first.
Find me
More about my work is at bagisetti.com.