
Nicolas Dupuis
Building deterministic data tools for AI agents turning messy scraper output into clean, exact, token efficient reports.
Joined July 2026
ACTOR STATS
2 public Actors
4 total users
1 monthly user
>99% runs succeeded
I build deterministic middleware for AI agents — tools that sit between raw scraper output and your agent, doing the heavy lifting that LLMs do poorly.
The problem I solve: scrapers return huge, messy datasets (80+ fields per product, duplicates, noise). Agents drown in tokens, hit context limits, and hallucinate numbers when they try to calculate.
My approach: 100% deterministic Python. No LLM in the pipeline. Exact figures, never hallucinated. Massive token reduction (90–99%).
⚒️ What I build: data cleaning, deduplication, price analysis, anomaly & fake-discount detection — all agent-first, all exact.
🚀 First tool: E-commerce Data Cleaner & Analyzer — turns Amazon, AliExpress, eBay or Temu scraper output into a compact, AI-ready report.
🫱🏻🫲🏽 Open to collaboration and feedback.