Overview
Own the data and modeling side of a small, dedicated bot-detection effort inside Hugging Face's Trust & Safety team, turning large volumes of traffic, account and behavioral data into decisions about what is human, automated, or abusive.
What you'll do
- Build and run detection pipelines on large-scale traffic and activity data (requests, sessions, accounts, repo interactions, API usage)
- Develop and deploy detection models and heuristics: anomaly detection, clustering of coordinated activity, graph-based approaches, supervised classifiers where labels exist
- Define features from client and network signals together with the signals engineer, and measure how robust they are as attackers adapt
- Build the feedback loops: labeling, evaluation, false-positive monitoring, dashboards, alerting
- Investigate live campaigns end-to-end (fake account waves, metric inflation, large-scale scraping) and turn what you learn into durable detections
- Keep things proportionate and privacy-respecting, and minimize impact on legitimate users and automated-but-legitimate workloads (CI, research crawlers, integrations)
What you'll need
- Strong experience applying data science or ML to adversarial problems in production: bot detection, fraud, abuse, spam, security analytics or similar
- Fluent in Python and SQL
- Comfortable with large-scale data tooling (Spark, ClickHouse, BigQuery, or similar)
- Experience taking models or detection logic from exploration to production and keeping them working over time
- Understanding of web traffic and network data sufficient to reason about what a signal really means and how it can be spoofed
- Thinks in terms of precision/recall trade-offs, attacker adaptation and operational cost, not just offline metrics
- Pragmatic, autonomous, and enjoys working closely with one or two people in a fast-moving team
Nice to have
- Published work on bot detection, web measurement or adversarial ML (a PhD is welcome but not required, as long as you want your work running in prod)
- Experience with graph analytics or streaming detection
- Familiarity with the ML/open-source ecosystem and the Hugging Face Hub
Details
- Remote, EMEA
- Hugging Face has office spaces in NYC and Paris; remote employees have the opportunity to visit the offices
- Flexible working hours and remote options
Read the full description and apply on the company’s own careers page.