About Me
I’m Francisco — a machine learning engineer and applied statistician based in Chicago.
My work sits at the intersection of statistical theory and production systems. I’ve spent the last decade building ML in environments where the math has to work and the code has to ship: embedded microcontrollers with 512 kB of memory, real-time sensor pipelines for autonomous vehicles, and two-sided marketplace models at Lyft scale.
Currently I’m a Staff ML Engineer at Lyft, working on marketplace optimisation — price elasticity, incentive targeting, driver supply forecasting, and graph-based lookalike models over the Lyft ride network. This work powers Lyft Ads.
Before that I was at Renesas Electronics / Reality AI deploying compressed neural networks (INT8 quantisation, low-rank factorisation) onto automotive-grade hardware. And before that, BCG and Deloitte, building production C++ inference pipelines for industrial clients.
I hold an M.S. in Computational Economics and Theoretical Computer Science from the University of Chicago.
This blog is where I write about things I find interesting enough to work through in detail: signal processing, deep learning theory, causal inference, reinforcement learning, the occasional book, and whatever I happen to be thinking about. The posts tend to be long and mathematical. I write them mostly for myself — to force precision — but I hope they’re useful to others.
If you want the full picture, the CV is here.
You can reach me at Mendes.franciscoromaldo@gmail.com or find me on GitHub and LinkedIn.
Popular Posts
- A Manual Implementation of Quantization in PyTorch — Single Layer
- Bayesian Statistics: A/B Testing, Thompson Sampling, Recommendation Engines and more
- Soft Actor Critic (Visualized) Part 2: Lunar Lander Example from Scratch in Torch
- Matching MATLAB's resample function in Python
- Signaling, Skills, and Intellectual Health in the Age of AI
- The Management Consulting Playbook for AB Testing
