AboutSan Francisco

I build LLM systems end to end

I am a software engineer working on applied AI. I build the systems that sit around machine learning models: the pipelines that feed them, the agents and products built on them, the evaluation that checks their output, and the infrastructure that keeps them running. Robotics is where I trained; applied AI is where I work.

Where the instinct comes from

My background is in robotics. I spent my master's at the University of Maryland working on perception and control, where a system that is confidently wrong does not fail quietly. That experience shaped how I approach machine learning: I am less interested in whether a system produces an answer than in whether we can tell when the answer is wrong.

The coursework is still on GitHub and I have kept it on this site, honestly labelled: a TurtleBot perception stack whose horizon comes from a hand-written RANSAC vote, planners whose termination rules can be shown sound or unsound, a YOLO stack on a live ROS 2 topic. None of it is polished software. All of it is where the habit of asking "how would we know?" started.

How I work

I work across the full stack, largely because the problems worth solving rarely stay in one layer. A slow page turns out to be a missing database index; a model that appears accurate turns out to be a pipeline that stopped running. Diagnosis is the part of the work I find most rewarding.

I place particular value on the engineering that is easy to skip: tests that genuinely fail when the code breaks, alerting that fires when it should, and migrations that can be rolled back safely.

The standard I hold tests to

A test only counts once a deliberate defect in the source proves it can fail, and the defect itself is verified as landed. Tests that pass against broken code are worse than no tests: they transfer confidence without earning it.

The path

  1. Jul 2019 to Jun 2023 SRM Institute of Science and Technology B.Tech, Mechatronics, robotics specialization Control systems, embedded systems, sensors and signal conditioning.
  2. Jul 2023 to May 2025 University of Maryland, College Park M.Eng, Robotics Perception for autonomous robots, planning, control, reinforcement learning. Robotics lab
  3. Sep 2025 to May 2026 Handshake AI AI Data and Evaluation Calibrated rubric evaluation of frontier image models; the human side of the evaluation pipelines I now automate. Case study
  4. Apr 2026 to present Verita AI Software Engineer, San Francisco Production LLM systems: audit pipelines, an annotation platform core, a rubric verifier, SOC 2 readiness, security tooling. Case study

Sources Résumé Verified 3 Sep 2026