from the author
Miguel
Chitiva Diaz
ML-Data Scientist
& resident overfitter
& resident overfitter
“calm, drama-free MLOps”
— the whole thesis, really
§ 01 — hello
Nice to meet you.
I wrangle messy, real-world data into products your users actually love
— from classical ML to agentic systems — delivered with
calm, drama-free MLOps. This is where I write it down so I
don’t forget how.
§ 02 — what I do
What I do
Ten-ish years of shipping ML at scale — classical pipelines, deep learning, on-device inference, and more recently, agentic systems. Tech lead by day, still a student of the craft at night.
#edge-AItiny sorcery to run big brains on small chips
#agentslocal LLMs with manners
#generativediffusion, VAEs, transformers — the usual suspects
#mlopsboring in production on purpose
#researchpapers, read and re-implemented
#data-engthe bit nobody thanks you for
§ 03 — currently
Currently
Right now:
tiny sorcery to run big brains on small chips
— quantization, distillation, and the gnarly art of squeezing
transformer-class models into edge silicon. Plus a side quest in
local LLMs with manners: retrieval, tool use, and
eval harnesses that don’t lie to you.
● STATUS: shipping
● LOCATION: latent space
§ 04 — say hi