Sudhanva

About

Sudhanva Narayana.

Senior ML Engineer at Montai Therapeutics , based in the San Francisco Bay Area and working remotely. I build production ML infrastructure on Kubernetes.

At Montai I own end-to-end ML infrastructure: batch prediction pipelines processing 1B+ rows, multi-GPU build systems, auto-scaling Kubernetes clusters with Ray and Flyte, and the observability + cost-control that keeps it all honest in production.

Before Montai, I interned at Autodesk on a real-time event pipeline serving a transformer for next-click prediction, and earlier built multi-regional ML infrastructure for geo-spatial analytics at Pixxel, India's leading private space-tech company. Earlier still: ML at Initiable Intelligence, and software engineering at DCT Academy.

This site is where I write up the parts of the work worth writing up, long-form notes on production ML, infrastructure, vector databases, Kubernetes (managed and bare-metal), and the engineering decisions behind all of it.

Experience

  1. Jul 2023, present

    Senior ML Engineer Montai Therapeutics

    San Francisco Bay Area, CA · Remote

    • Built batch prediction pipelines processing 1B+ rows across TensorFlow models in under 3 hours, accelerating drug-candidate evaluation.
    • Owned end-to-end ML infrastructure: auto-scaling Kubernetes with Ray and Flyte, observability (Prometheus, Grafana, alerting), $50k+/yr cloud-cost reduction.
    • Architected a multi-GPU build system that halved model deployment time; CI/CD + automated tuning save ~10 hours/week.
    • Shipped internal apps enabling cross-functional teams to run models on-demand with experiment tracking and model versioning.

  2. May 2022. Jul 2022

    Machine Learning Engineer Intern Autodesk

    San Francisco, CA

    • Deployed a high-throughput transformer on a real-time event pipeline, analysing 1M+ daily user interactions for next-click prediction.

  3. Jun 2020. May 2021

    Machine Learning Engineer Pixxel

    Bengaluru, India

    • Built multi-regional ML infrastructure for geo-spatial analytics, 75% inference-efficiency gain, 50% map-render latency reduction.
    • ETL + ML pipelines processing 1TB+/day; pipeline failures from schema issues cut by 30%.
    • Architected a GPU queue with automatic A/B testing for regional model selection, ~$50k/yr saved.

  4. May 2019. May 2020

    Machine Learning Engineer Initiable Intelligence

    Bengaluru, India

    • Led a team of 4 to build an NLP-powered chatbot API enabling 5-minute integration for static sites; adopted by multiple startups.

  5. Jan 2017. May 2019

    Software Engineer DCT Academy

    Bengaluru, India

    • Built a task recommendation engine with automated assignment matching, 60% student engagement lift.
    • Built a classroom-management tool with auto-grading and student clustering, ~2 hours/day saved for teaching staff.

Education

  1. Sep 2021. May 2023

    Master of Science, Artificial Intelligence

    Northeastern University

  2. Jun 2013. Nov 2016

    Bachelor of Science, Computer Science

    PES University