Chandrahas Aroori
AI @ Salesforce · PyTorch & Kernel Lover · ex-Microsoft, ex-Nvidia
Believer in the power of AI. Fundamentals matter. Domain Expertise matters. Build with me.
Into the intersection of AI and the real world. Author of TorchLeet (2k+ stars) and AdKit-MCP
Follow my unfiltered thoughts on Twitter/X →
Leetcode for PyTorch — a collection of hands-on exercises to learn PyTorch by solving problems. 1,958+ stars.
A cookbook to start building with LLMs. Practical notebooks covering fine-tuning, RAG, and prompt engineering. 112+ stars.
A collection of principles I've picked up from building and using command-line tools over the years.
The obligatory first post — what this site is, why I built it, and what to expect next.
Building AgentForce step by step at Einstein Studio.
Multi-model systems with Llama 8B and Phi 2.7B. Achieved 35% model size reduction and 70% inference speed improvement through fine-tuning.
Reinforcement learning and Bayesian optimization for ADAS resource allocation.
Built ML features: Picture to Tasks (78% accuracy), Discover Feed (20% engagement increase), Autosuggest Tasks, Autocategorize Tasks (35% manual effort reduction).
Enhanced HDMI I2C graphics driver functions and GPU power performance testing utilities.