Senior AI Engineer @ Meta · Previously Amazon · ~10 years across AI/ML and large-scale platforms
Senior AI Engineer at Meta, working on applied AI and ML systems that serve at scale. Before that, Amazon, where I built large-scale platforms. Roughly a decade across machine learning, distributed systems, and platform engineering.
My day to day lives at the intersection of LLMs, agentic AI, and production infrastructure: training and inference pipelines, agents that do real work, and the platforms underneath them. Outside of work I teach, mentor, and speak in the AI/ML community.
A few things from my repos, grouped by theme.
Agentic AI and LLM apps
- career-ops: AI-powered job search system built on Claude Code. 14 skill modes, a Go dashboard, PDF generation, batch processing. repo
- agentic-contact-center: agentic AI for contact center automation (Python). repo
- adk-samples: sample agents built with the Agent Development Kit. repo
- ai-engineering-from-scratch: notes and builds. Learn it, build it, ship it. repo
Models and ML research
- Megatron-LM: experiments in training transformer models at scale. repo
- minGPT: minimal PyTorch re-implementation of GPT training, mostly a learning exercise. repo
- LLaVA-Med: language-and-vision assistant for biomedicine. repo
- alphafold3: AlphaFold 3 inference pipeline. repo
Computer vision
- Face mask detection: computer vision + deep learning with OpenCV and TensorFlow/Keras. repo
- Brain tumor MRI segmentation: tumor detection and extraction from brain MRI scans (Python). repo
Python, TypeScript, Go, Java, C++, Kotlin, JavaScript. PyTorch, TensorFlow/Keras, transformers, LLM agents (ADK, Claude Code), OpenCV. AWS, Docker, Terraform, distributed systems, CI/CD.
Won the "Roast My PR" hackathon with a TypeScript + Python build (frontend / backend). I also judge hackathons on Devpost and MLH.
- Instructor at Newton School: full-stack, advanced DSA, and advanced Java batches. My course notes are some of my most starred repos.
- Masterclasses at Tutort Academy.
- Technical organiser at GDG Singapore.
- Mentor at Manifest/EAC.
I mentor 1:1 at multiple places, in this order:
- Topmate: mentorship on careers, interviews, and ML system design. topmate.io/idevangsharma
- Exponent: coaching on coding, ML system design, behavioral rounds, and career growth. tryexponent.com/coach/Devang-Sharma
- MentorCruise: long-term mentorship for engineers. mentorcruise.com/mentor/devangsharma
- MeetaPro: meetapro.com/provider/idevangsharma
Spoken at Silicon Valley AI Hub and ClueCon.
- Best resources for a programmer: a curated compilation.
- system-design-primer: designing large-scale systems, with Anki flashcards.
- system-design-important and low-level-design-primer: more system design prep.
- Full-stack course notes: from my Newton School March batch.
- Interview prep collections: JavaScript, React, Java, and company-wise problem sets.
Google Scholar · Semantic Scholar · ORCID · Wikidata
