Tomas D'Avola0%
Skip to the work
Tomas D'Avola
Tomas D'Avola

About

I build AI products and the infrastructure behind them: machine learning systems, large-scale data platforms and developer tools, at Apple, Phia and FileAI.

Based
Manhattan, NY
Origin
🇦🇷 🇸🇬
Languages
EN / ES / 中文
Available
Jan 2027

Experience

Software Engineer, Manhattan, NYJAN 2026 – JUL 2026
0.90auc

led an end-to-end predicted-LTV system forecasting 90-day value from a user's first day, on 1.1B+ rows.

580K+users clustered

built the segmentation behind app personalisation: 6 segments, 91% reproducible on a held-out month.

450+bug tickets

resolved by autonomous LLM agents built to auto-fix scraper failures across 3,000+ retailer sites.

$400Kper month

built the 6+ dashboards (ROAS, CAC, LTV, OKRs) that ad-spend decisions run on.

20K+ads ranked

built a competitor-intelligence platform across 1,800+ companies, on language, video and object detection.

DATAMLINFRAFRONT

Projects

1st
HackMIT, SEP 2026

Hear only the person you're looking at.

3×speech ratio

built directional noise cancelling on four commodity mics, reaching 80% of the theoretical limit.

46%clearer speech

made real recordings more intelligible by computed score, +0.19 STOI.

~50msend to end

built SRP-PHAT localization at 21 fps to 1–5°, steered by an ESP32 head tracker over WiFi.

MVDRSRP-PHATDeep Neural NetESP32Python