I'm Manas Nanivadekar, a research engineer at IISc Bangalore's LEAP Lab working on speech and audio ML: unsupervised ASR, AI-generated voice detection across Indian languages, and speaker diarization. I'm studying Data Science at IIT Madras and Computer Engineering at Mumbai University, and spent time building cloud infrastructure at a London startup and pricing derivatives at a quant firm. I hold six AWS certifications and have taught as an AWS Authorized Instructor .

I like building things from first principles, whether that means diffusion models from the math up, LSTMs and optimizers from scratch, or distributed crawlers in Go across EC2 fleets. I'm currently writing Thinking in Transformers, a technical primer meant to teach people to think in the design space of LLMs, not just understand the mechanics. I care more about depth than breadth, and I'd rather wrestle with the hard thing than ship the obvious one.