rishi
sde @ intersystems, previously cs @ johns hopkins. i enjoy making language models faster, cheaper, and easier to trust.
🔬 research
- making llms lazy in a good way — think hard only when it matters
- hiding your face on camera without ruining the footage
- "hey siri" but it actually works at a concert
🛠️ side projects
- finding where "aligned" models secretly get worse
- what is my gpu doing between tokens?
- what if ai agents could mentor each other?
- squeezing more speed out of LLMs on a laptop gpu
💼 experience
- 🏗️ currently @ intersystems
- 🍃 summer 2025 @ jhu dsai
- 📚 fall 2024 @ jhu clsp
- 🔨 summer 2023 @ mastek
🎲 fun
- 🧗 bouldering
- ♞ chess
- 🏋️ lifting
- 🥐 baking