About

I'm Dhruvkumar Patel — a Data Scientist and ML Engineer finishing my M.Tech at IIIT Delhi (graduating June 2026). I currently work at Myntra on the Ads Ranking team, building production ML infrastructure for CTR optimization, and research at MIDAS Lab on multimodal LLMs and mixture-of-agents frameworks.

My path into ML wasn't a straight line. I started with web projects and hackathons — winning SSIP-22 and reaching the national finals of KAVACH-23 — before realizing that the problems I found most interesting all sat at the intersection of scale, data, and inference. That's what pushed me toward graduate research and applied ML engineering.

What I care about: systems that work in production, not just on benchmarks. Most of my time is spent on the unglamorous parts — feature pipelines, experiment reproducibility, evaluation harnesses — because that's where real models fail.

Current focus

  • At Myntra — re-architecting the AdsRank platform on Databricks/PySpark and automating ML experiment workflows with MCP agents.
  • At MIDAS Lab — building async batch inference pipelines for multimodal LLM annotation and improving small LLM generation via Mixture of Refinement Agents.
  • Writing — this blog and technical deep-dives on ML infrastructure patterns.

Working with me

I prefer async over sync, direct feedback over diplomatically vague feedback, and concrete problem statements over open-ended explorations. I'm best at the interface between research and engineering — taking a model or technique from paper to deployed system. I'm looking for full-time roles as a Data Scientist or ML Engineer starting July 2026.

If you want to collaborate, recruit, or just talk ML infrastructure — reach out.

Personal stack

  • VS Code — Primary code editor, customized for efficiency.
  • Neovim — For fast terminal-based editing and scripting.
  • Notion — Project management, note-taking, and research organization.
  • PenPot — UI/UX design and visual brainstorming.
  • iTerm2 — Versatile and powerful terminal for macOS.
  • Databricks — Primary platform for ML experimentation and feature engineering.

This site is built with Next.js 15, Tailwind CSS, and MDX. The name “Proemio” — an introduction or preface — was what this page used to be called. It has since grown into something more honest.