Sanjan Das

Building AI-powered products — from LLM agents to scalable inference systems.

I like building things with AI. My work sits at the boundary of machine learning research and production software — taking ideas from papers and turning them into systems that actually work at scale.

Most recently I was at Amazon, where I helped launch and scale Bedrock Batch Inference to 100B tokens/day. Before that I spent time at Carnegie Mellon working on NLP problems like factuality in summarization and transformer optimization for edge devices.

Now I'm building Dyno Apps, a no-code mobile app platform powered by AI coding agents. I'm interested in how LLMs can go beyond chat interfaces and become reliable tools for building software.

Education

  • Carnegie Mellon University — M.S. in Artificial Intelligence and Innovation, 2023
  • University of Cambridge — B.A. Bioengineering, M.Eng. Computer and Information Engineering, 2021

Get in Touch

Featured Projects

View all →

Authentication Recommendation System

2023

Carnegie Mellon University

ML-based authentication recommendation system for Indonesian bank BTPN, improving fraud detection by 95% over existing rule-based systems.

Python · PyTorch · Machine Learning

DreamControl-3D: ControlNet for Text-to-3D

2023

Carnegie Mellon University — 16-825 Learning for 3D

Added sketch-based control to text-to-3D generation by integrating ControlNet with DreamFusion, enabling users to guide 3D model creation with simple scribbles.

GitHub Python · PyTorch · ControlNet · NeRF · Stable Diffusion · Deep Learning

Latest Post

All posts →

Hello World

A first post to kick things off.

meta