Hi, I'm Mrinal
Welcome to my corner of the internet where I dump my thoughts.
$ echo "The world is a level 99 game with no tutorial. If you know, you know."
About Me
I'm a Research Engineer currently exploring the territory of reinforcement learning and continual reasoning at Dyssonance AI. My superpower is that I look at a problem through an adversarial lens. My interests span across interpretability and misalignment of AI systems.
I occasionally mentor students navigating their first steps in ML or early career decisions in tech, it's my way of giving back to the communities I come from. If you're part of the UC or IIT system and want to chat about machine learning, research, or breaking into the industry, feel free to reach out at {first_name}.{last_name}07{at}gmail.com.
A Contrarian Take on the Alignment Problem
Why aligning superintelligence to 'human values' might be aiming at the wrong target, and what it could mean to build the values in before the knowledge.
Read the essay →The Alignment Problem and Shoggoth Meme
First in the series about understanding AI misalignment. Meme are getting real!!
Read the essay →Self-Evolving Search Agents: How LLMs Learn Without Training Data
Exploring Dr. Zero's framework where LLM agents bootstrap their own training data through self-play, enabling continuous improvement without human annotation.
Read the essay →Learning to Adapt in Test-Time (Titans/MIRAS)
A deep dive into Titans and MIRAS architectures that enable LLMs to memorize and adapt at inference time using neural memory modules.
Read the essay →Towards Infinite Context: How LLMs Are Breaking the Context Limit
A comprehensive guide to extending LLM context windows through position encodings, efficient attention, and memory augmented architectures.
Read the essay →Attention That You Probably Didnt Know Existed!!
From sparse patterns to linear attention and state space models exploring the zoo of efficient attention mechanisms that go beyond vanilla transformers.
Read the essay →|ψ⟩ = (|Max Verstappen⟩ + |Ferrari strategy⟩) / √2