Nishant Modi

Management Engineering student at the University of Waterloo. Previously, I spent time as a Data Science & AI Intern at RBC where I built tools for fraud detection, and as an ML Research Intern at AquaShield AI (YC P26), where I worked on applied ML evals for water leak detection.

Exploring applied research, evals, and inference. You can reach me on X, GitHub, LinkedIn, or email.

Writing

Content-Adaptive Sparse Attention Kernel for Long-Context Inference Aug 2026

A novel Triton kernel that picks which tokens to attend to at runtime, computing 3.9% of dense attention's FLOPs at 32k context.
Temporal Memory Architecture in Multi-Agent Loops Jul 2026

As multi-agent loop engineering matures, memory architecture becomes the critical failure point for most teams.

Thoughts

Close your loops Aug 2026

Here's how I reduce burnout and stay present.

🎧

⛹️‍♂️

📺