Leak It: Per-Document Extraction Beyond Aggregate Membership Inference
arXiv preprint, 2026
Senior Machine Learning Engineer at Uber.
I'm a Senior Machine Learning Engineer at Uber. Previously I was a Staff Machine Learning Engineer at Upwork, working on Uma, Upwork's mindful AI, under the Algorithms & Research team. My focus is fine-tuning small language models for function calling and search-query understanding under tight latency budgets.
Before that I was a research assistant at MIT CSAIL working on protein language models, a genomics data scientist at Mendelics (where a variant prediction model I built ended up in Brazil's national newborn screening program), and a data scientist at OLX and Shell.
I formally studied Medicine before moving fully into machine learning, and I've kept medical sciences as a serious intellectual interest. That background still shapes how I think about biomedical AI, diagnostics, and useful health systems. I'm currently pursuing an MSc in Health Informatics at Karolinska Institutet to keep those two tracks talking to each other.
I build side projects: focusy.co, enhancv.me, kannel.ai, peakframe.ai, and adhd-screening.org.
For the medical thread of all this, the questions I keep coming back to in biomedical AI, diagnostics, and how health systems actually work, there is a separate /med page.
Reach out at maricatovictor [at] gmail.com. My CV is also available as a PDF (LaTeX source here).
arXiv preprint, 2026
SSRN preprint; poster accepted, XLIII Congresso Brasileiro de Psiquiatria (CBP) 2026, 2026
EMNLP (Industry Track), 2025