Senior Software Engineer-ML Platform
RobloxSan Mateo, CA, United StatesOn-Site
Needs a free Superjobs account. You'll finish on Roblox's site.
Skill score
Summary
The engineer builds and operates the data and serving foundation for user-level safety understanding, including a safety user feature store, user embeddings, and risk scores, and takes them from offline training to low-latency production serving and streaming updates. This role suits a senior engineer with 5+ years of production ML platform or data infrastructure experience who enjoys owning reliability, governance, and end-to-end delivery in an ambiguous safety domain.
- Function
- Data Science
- Estimated salary
- Est. $155k–232k/yr
- Estimated for a senior role, ~5 yrs experience, United States. Not published by the employer.
- Work mode and location
- On-Site, San Mateo, CA, United States
- Experience
- 5+ yrs
Job description
Overview
As a Senior Software Engineer on the User Behavior ML Platform, you will build the data and serving foundation behind Roblox's user-level safety understanding, so that every safety system can see who a user is, not just what a single piece of content looks like.
You will own the core ML platform for a new capability at Roblox: a safety user feature store, user embeddings, and risk scores that power personalized consequences, proactive detection, and early education. This is a rare opportunity to build foundational, high-scale ML systems that shape how Roblox understands and responds to risk across hundreds of millions of users.
You will feel a deep sense of responsibility building systems that directly affect millions of users, especially minors, on one of the world's largest platforms. Your work will reduce harm, improve fairness, and help ensure Roblox remains a place where creativity and community can thrive safely.
Roles and responsibilities
- Build the Safety User Feature Store: Design and operate the pipelines that turn violations, reports, appeals, filter activity, and engagement signals into reliable, well-documented user features for training and serving.
- Serve User Embeddings and Risk Scores: Take user embeddings and risk models from offline training to production, including low-latency online lookup and similar-user search for detection teams.
- Build the Real-Time Path: Extend batch pipelines to streaming, so fresh behavioral signals reach models and downstream safety systems within minutes, not days.
- Enable ML Teams to Move Fast: Partner with MLEs and Data Scientists to provide self-serve tooling for feature discovery, training data generation, offline evaluation, and A/B experiment hooks.
- Own Data Quality and Governance: Build monitoring for feature freshness, drift, lineage, and ownership, so teams across Safety can trust and reuse signals.
- Design for Trust and Auditability: Build systems that support explainability, fairness review, and appeals correction, since these signals directly shape consequences for millions of users.
- Ship Code, Not Just Designs: Own end-to-end delivery, from system design through migration, testing, production operation, and oncall.
Skills required
Top skills
- ML platform
- Batch and streaming pipelines
- Spark
- Flink
All skills (9)
- Kafka
- Feature stores
- Low-latency serving
- gRPC
- Key-value storage
- Python
- Go or Java
- Production operations and oncall
- Trust and Safety
Qualifications
- 5+ years of experience building and operating production ML platforms, data infrastructure, or large-scale backend systems.
- Strong experience with batch and streaming data pipelines (e.g., Spark, Flink, Kafka) feeding feature stores or similar ingestion layers.
- Experience designing low-latency serving systems: service APIs (RPC/gRPC), key-value storage, caching, and high-throughput read paths.
- Proficiency in Python and a statically-typed backend language (Go, Java, or similar).
- Solid production operations fundamentals: monitoring, alerting, oncall, and incident response for systems with real user-facing impact.
- Comfort partnering with ML engineers, with a working understanding of the model lifecycle: training data generation, offline evaluation, deployment, and drift monitoring.
Nice to have
- Experience in Trust & Safety, content moderation, user detection, or other adversarial domains
- Experience with user-level modeling or user understanding
- Experience with embeddings and vector similarity search in production
- Familiarity with feature store platforms and signal governance
- Experience building agentic workflows or tool interfaces for LLM agents (e.g., MCP)
Other details
- A Systems Thinker: capable of designing platforms that ML and detection teams build on, not just point solutions.
- A Pragmatic Builder: scrappy and impact-oriented, comfortable turning scattered, undocumented signals into clean, well-owned data products.
- Comfortable with Ambiguity: this space is still being defined, and you will help set the standard rather than follow an existing one.
- A Reliability-Minded Engineer: you treat data quality, testing, migrations, and oncall as first-class parts of the job.
- A Strong Collaborator: you work effectively across ML, Policy, Product, and Detection teams, and make the people around you faster.
- Committed to Ethical, Responsible Engineering: dedicated to building systems that are fair, auditable, and operate with the utmost responsibility toward user safety.
Compensation
For roles that are based at our headquarters in San Mateo, CA: The starting base pay for this position is as shown below. The actual base pay is dependent upon a variety of job-related factors such as professional background, training, work experience, location, business needs and market demand. Therefore, in some circumstances, the actual salary could fall outside of this expected range. This pay range is subject to change and may be modified in the future. All full-time employees are also eligible for equity compensation and for benefits as described on this page.
Annual Salary Range
$196,750—$243,290 USD
Benefits
Roles that are based in an office are onsite Tuesday, Wednesday, and Thursday, with optional presence on Monday and Friday (unless otherwise noted).
Company info
About Roblox
Roblox is an platform where users imagine, create, and play together in immersive, user-generated 3D worlds. Millions of people use it to build and share experiences and to play them with others.