Staff / Principal Research Engineer - Data Flywheel & ML Data Infrastructure

Neolix; Sunnyvale (CA); On-site; USD 180,000 - 240,000; Full time

Benefits offered by this job

Equity

Premium health insurance

401(k)

Job summary

NeolixUS in Silicon Valley is seeking a hands-on technical leader to build the data flywheel that connects fleet experience to model improvement. You will own the mining-to-training data platform and enable scalable, reproducible datasets for driving models.

This role combines distributed systems, applied ML and dataset science to accelerate model improvement per unit of usable experience, not merely storage.

Qualifications

• Must have hands-on ownership of scalable distributed data systems or ML data infrastructure.
• Experience with multimodal or sequential datasets and ML workflows.
• Proven leadership across ML, infrastructure and production teams.

Responsibilities

• Architect a petabyte-scale discovery platform across video, images, crops, sensor context, model outputs and metadata.
• Define cloud-side trigger framework and prioritize informative long-tail scenarios within bandwidth/storage limits.
• Create version-controlled pipelines for discovery, collection, annotation, QA, deduplication, curation and publication.
• Deliver deterministic, high-throughput training data mixtures for end-to-end driving models.
• Turn model weaknesses into mining priorities and prove improvements with controlled comparisons.
• Set architecture and interfaces across regional data teams and uphold privacy and data locality.

Skills

Python

C++

Go

SQL

Data modeling

Job description

Neolix US is the Silicon Valley Physical AI research team of Neolix, the global leader in L4 autonomous delivery. Neolix operates 30,000+ autonomous logistics vehicles in 300+ cities across 15+ countries, with ~200 million autonomous kilometers driven on public roads.

About the role

We are looking for a hands-on technical leader to build the data flywheel that connects fleet experience to model improvement. You will own the mining-to-training data platform: the systems that discover valuable scenarios, capture the right evidence, turn raw multimodal data into reproducible training datasets, and make those datasets usable by our end-to-end driving models.

This is not a storage or ETL role. You will work where distributed systems, applied ML and dataset science meet. Success means faster, more reliable model improvement per unit of usable experience, not simply more stored petabytes.

Must-have: hands-on experience at a leading autonomous-driving company (e.g., Waymo, Tesla Autopilot, Wayve, Momenta, XPeng, Cruise or Nuro). General tech experience alone does not qualify.

What you'll own

• Multimodal search and hard-case discovery: architect a petabyte-scale discovery platform across video, images, object crops, sensor context, model outputs and episode metadata, combining embeddings, vector search and SQL/metadata filtering for text-to-video search, example-based retrieval and scenario exploration
• Fleet mining and selective capture: define trigger semantics and the cloud-side trigger framework, and partner with vehicle engineering to prioritize informative long-tail scenarios, interventions and outcomes within bandwidth, storage and safety limits
• Reproducible data-engine workflows: version-controlled pipelines for discovery, collection, annotation, QA, deduplication, curation and dataset publication, with end-to-end lineage, observability and self-service tools for researchers
• Feature materialization and training data access: a versioned layer for multimodal features and curated datasets that delivers deterministic, high-throughput training mixtures for end-to-end driving models
• Learning-value measurement: work with the end-to-end driving and Data Recipe owners to turn model weaknesses into mining priorities, and prove with controlled comparisons that targeted data beats matched random additions
• Technical leadership: set architecture and interfaces across regional data teams, mentor engineers, and uphold privacy, dataset rights and data-locality requirements

Qualifications

• Required: experience at a leading autonomous-driving company; general tech experience alone does not qualify
• Significant hands-on ownership of scalable distributed data systems, ML data infrastructure or production data platforms, from architecture to operation
• Strong Python plus a systems language such as C++ or Go; fluent SQL, data modeling and performance debugging
• Experience with large-scale multimodal or sequential datasets and ML workflows, including feature extraction, dataset versioning, training data access and evaluation splits
• Ability to connect data quality, distribution and sampling decisions to model experiments, not just pipeline throughput
• Proven technical leadership across ML, infrastructure and production teams, taking complex systems from design to dependable use
• Sound judgment on reliability, cost, privacy and security, and on designing within real vehicle resource constraints

Preferred

• Autonomous-driving or robotics data flywheels, fleet-trigger systems, expert interventions or task-outcome capture
• Large-scale vector retrieval, hybrid search, video understanding, multimodal embeddings or VLM/batch inference platforms
• Active learning, uncertainty-aware selection, hard-example mining, auto-labeling and long-tail scenario taxonomies
• High-throughput feature stores, distributed preprocessing and deterministic data delivery for large-model training
• Edge-cloud systems with controlled configuration rollout, limited bandwidth and constrained onboard compute
• Equity: generous founding-team equity package
• Benefits: premium health, dental and vision insurance, 401(k), and comprehensive Silicon Valley tech perks

You will help define how a Physical AI organization learns from real-world operation: discover the right experience, make it reproducible, and show that it changes autonomous behavior.