Senior Machine Learning Engineer, User Behavior

Roblox Roblox · Consumer · San Mateo, CA · Software Engineering

Senior Machine Learning Engineer focused on building proactive detection and behavioral intelligence systems to identify users at risk of violating policies before escalation. The role involves owning the ML foundation from scratch, including feature stores, training pipelines, and evaluation infrastructure, and shipping end-to-end ML solutions integrated with the main stack. This is a consumer-focused role building agentic systems for risk assessment and intervention.

What you'd actually do

  1. Define and lead the multi-year ML vision, architectural strategy, and execution that identify casual violators early before they escalate, and operate reliably under real-world policy and precision constraints.
  2. Develop and maintain production ML models that translate user behavioral signals into actionable risk intelligence, enabling earlier and more targeted interventions.
  3. Design the user feature store, training pipelines, and model evaluation infrastructure that underpin behavioral ML at Roblox.
  4. Partner with XFN stakeholders to ensure A/B tests on intervention variants have proper random assignment and instrumentation for future causal model training. This data cannot be retroactively collected.
  5. You will not just model; you will build. You will work cross-functionally to construct datasets from scratch where none exist, build labeling pipelines, and ship solutions to various product surfaces.

Skills

Required

  • 5+ years of experience building and shipping production ML systems at scale
  • ML classification at scale (traditional ML, sequence modeling, LLM-based approaches)
  • model evaluation on real-world data
  • designing and maintaining feature pipelines and feature stores for behavioral or event-driven data
  • building embedding-based retrieval systems (ANN / vector search) in production
  • building and operating end-to-end ML pipelines (training, offline evaluation, A/B experiment hooks, production serving infrastructure)
  • ML experiment design (training data validity, label quality, recall/precision tradeoffs, causal inference fundamentals)

Nice to have

  • Trust & Safety experience
  • user behavior modeling
  • user understanding (e.g., churn prediction, engagement modeling, risk scoring)
  • uplift/causal ML familiarity
  • sequence models on user-event data

What the JD emphasized

  • own the ML foundation for a new capability at Roblox
  • build the ML foundation from scratch
  • This data cannot be retroactively collected
  • ship solutions to various product surfaces
  • ship code, not just models
  • own end-to-end delivery
  • integrate your models directly into the stack alongside SWE

Other signals

  • building proactive detection and behavioral intelligence systems
  • identifying users at risk of becoming chronic violators before they escalate
  • powering adaptive consequence systems
  • build a user-level behavioral intelligence capability
  • define and lead the multi-year ML vision, architectural strategy, and execution
  • develop and maintain production ML models that translate user behavioral signals into actionable risk intelligence
  • build the ML foundation from scratch: design the user feature store, training pipelines, and model evaluation infrastructure
  • instrument experiments for causal validity
  • drive end-to-end product development
  • ship code, not just models
  • own end-to-end delivery, building and operating ML pipelines from training and offline evaluation through production serving
  • integrate your models directly into the stack alongside SWE