Hello. I'm Abdur Rahim, a Senior AI Engineer based in Bangladesh, working remotely for ESAP, a Saudi Arabia-based company where I build real-time voice agents and conversational AI systems. My day-to-day involves frameworks like Pipecat, optimizing local LLM deployments, and making production-grade pipelines that actually hold up in the real world.

But my ambitions stretch far beyond engineering tasks.

I'm a problem solver at heart — I genuinely enjoy mathematics and the process of learning new things. Right now, I'm on a mission to master JAX. Not just use it, but think in it. I want to reach a point where I can translate any idea, paper, or algorithm directly into JAX or confidently migrate existing codebases. That kind of fluency with a tool is satisfying in a way that's hard to explain.

A lot of that learning ends up on GitHub. The biggest piece is mlx-tune: the Unsloth experience for Apple Silicon, fine-tuning LLMs and vision-language models natively on MLX. It recently passed 1,300 stars. The newest is mlx-dspark, which brings DeepSeek-style speculative decoding to the Mac. Around them are smaller pieces, like the speech servers I use in voice work and a Bangla keyboard for macOS. Most of them started as things I needed myself.

I'm also an active Kaggle competitor, one medal away from Master, with Grandmaster the goal after that. It's not just a title I'm chasing, it's the mastery it represents. Each competition is a puzzle, and I live for that.

What really pulls me in, though, is AGI. I'm drawn to computational neuroscience and what it can teach us about building smarter systems. I've been digging into the Thousand Brains Theory and following Numenta's Monty project, seeing it outperform ViT on some tasks got me excited. I've written about the ARC-AGI challenge, and I spend a lot of time thinking about fluid intelligence: that flexible, generalizable reasoning humans do effortlessly, but AI still struggles with.

My dream is to one day contribute meaningfully to this space — perhaps through systems inspired by how the brain actually works, whether that's hierarchical cortical models, spiking neural networks, or something we haven't discovered yet. I don't know exactly what form it will take, but I'm working toward it every day.

Beyond all this, I'm someone who genuinely loves reading papers. Not as homework, but as exploration. There's always something new being discovered, and staying at that edge is where I want to be.

Want to chat? Reach me at erahim3@gmail.com or on LinkedIn.