Roblox
New grad
PhD early-career systems ML role building real-time NPC inference and AI platform internals at Roblox
This is a PhD-early-career Senior MLE role split across two tracks: the Creator Services Machine Intelligence team building an NPC system that plays Roblox games and runs real-time inference at platform scale, and the ML Platform team building core AI infra (serving layer, model registry, pipeline orchestrator) and distributed LLM/recommender inference optimization. Expect deep systems-level ML interview questions on data pipelines, low-latency inference, and GPU optimization, calibrated to a strong PhD thesis rather than years of industry experience.
Step into this interview
Free · a live voice mock calibrated to this exact role
What this interview tests
- Real-time/low-latency ML inference at scale
- Distributed inference systems for LLMs and recommender models
- GPU-level optimization (continuous batching, speculative decoding, quantization)
- End-to-end ML pipeline design and data pipeline engineering
- ML platform components (serving layer, model registry, orchestration)
- Kubernetes and cloud infrastructure (AWS/Azure/GCP)
Common question themes
Design a data pipeline to collect 3D game state and player action data at scale
How would you architect a serving layer or model registry for hundreds of ML use cases
Walk through optimizing an inference engine to serve millions of QPS at low latency
Explain speculative decoding or quantization trade-offs on GPU hardware
How would you support real-time inference for 100 simultaneous autonomous agents
Describe your PhD thesis and how it maps to production ML systems work