Founding Engineer

Hub is a multimodal data lab for physical AI. We capture how humans actually move through the world: first-person video of everyday tasks, hands, homes, and environments, and turn it into the original training data that physical AI and robotics companies need and cannot get from the web.

Robots can't learn the physical world from the internet, because the internet never captured it. How a hand opens a jar, how a body moves through a kitchen, how a task actually unfolds: none of it exists as training data until a real person records it first-person. Hub is the pipeline that brings it in: not annotation, not scraping, not synthetic data. Original data, created end to end.

Headquartered in San Francisco and backed by Y Combinator (P26) with $2M raised, Hub's distributed contributor network spans hundreds of thousands of people across 150+ countries and 100+ languages. Contributors record original egocentric (first-person) video through our own platform, HubApp, verified through human-in-the-loop QA before delivery. We work directly with frontier AI labs, Fortune 500 companies, and top robotics companies, scaling multiple 7 to 8 figure contracts, to scope, collect, and deliver custom real-world datasets from scratch.

We're a small, fast-moving, deeply technical team of builders, data scientists, and ML engineers advancing embodied AGI through real-world data and research, solving the data ceiling that's become a critical bottleneck for physical AI. If you want to build the infrastructure that brings humanity's lived experience to the frontier of AI, we want to hear from you.

Robot foundation models learn physical behavior from human demonstration, and real-world data has become a critical bottleneck in embodied intelligence.

Hub provides real-world training data to frontier AI labs and robotics companies, scaling multiple 7 to 8 figure contracts in parallel. Our network of 150,000+ users captures egocentric human data across all continents, and we run collection, processing, QA and annotation in-house, end to end.

Hub's mission is to advance embodied AGI through real-world data and research.

We're assembling a unique team of exceptional and relentless individuals, willing to dedicate themselves fully to the achievement of this mission.

https://tr.ee/SwYtiM

As a founding engineer you own some of the most critical systems in the company.

• Multiple platforms with 150K+ users collecting data: B2C, B2B, mobile app, internal.
• Anti-fraud and automated QA systems. You own the quality of what we deliver every day.
• Heavy backend, keeping architecture and ops efficient as we scale to 1M+ users.
• Payout systems moving daily high volumes of money across 100+ countries.
• Live production serving multiple top 5 AI companies, hard deadlines and specs.

Graduates from top French engineering schools, ideally through classes préparatoires, 6 months to 3 years of experience. More experience is welcome when the track record justifies it. We look for outliers with outstanding achievements and grit.

• Exceptional individual achievement: elite rankings at competitions or concours, hackathons won, system you hacked, projects at real scale or revenue.
• You live for programming. Thousands of contributions on GitHub this year, active open-source participation (e.g frequent late nights commits on weekends on your own)
• You treat agentic engineering as a craft. You research the field daily and always run a SOTA customized harness.
• You're willing to sacrifice to be part of an extraordinary journey, working at the center of the embodied AGI frontier alongside the biggest AI labs.

• Small core team by design. Y Combinator + SF high-performance culture and pace.
• European founders, team across SF & Europe.
• High accountability. Features owned end to end: design, build, ship, support.
• Token-maxxing ethos. Uncapped frontier models, fleets of agents, shared skills and KB.
• Multiple features shipped per day, output of an org 50 times the size.
• Constant proactive communication. Slack, huddles, short standups. No black box.
• We hold a higher bar for ourselves than anyone could hold for us.

• Python, FastAPI, Postgres, React, Docker, AWS and S3, versioned REST APIs.
• Mobile: React Native and Expo, one codebase across iOS and Android, ARKit and ARCore under it for egocentric video capture. RGB+IMU sync, MCAP out.
• Multimodal pipelines at petabyte scale: ingestion, VLM QA and labeling, fraud detection, stereo depth and hand reconstruction, delivery in MCAP and LeRobot.
• We build our own custom platforms, not assembled SaaS: contributor app, commercial platform, mobile app, internal, payouts.

• Founding engineer in one of the fastest-growing Y Combinator startups.
• $80,000 to $100,000 yearly salary, stock options between 0.25% and 1%.
• Remote, based in France. Paris office opening soon.
• Direct work with the founders. Direct work with top AI labs.

Send us a short application, in your own words. We want to hear the thing you've done that you’re the most proud of, your professional handles (GitHub, links, etc), and the story behind your achievements.

If you want to spend the next years building the foundations of embodied AGI, working alongside the biggest labs, we want to hear from you.

At Hub, we build and operate distributed, multimodal infrastructure that orchestrates large-scale real-world data pipelines from collection to enterprise delivery.

Our engineering team builds and scales infrastructure across four core layers:

• Intelligent Distribution (HubApp): Algorithmic matching engines that dispatch collection tasks to a verified network of global contributors based on hardware, location, and language profiles.
• Automated ML Processing: Modality-specific pipelines for audio and egocentric video, including voice activity detection, denoising, speaker diarization, synchronization, motion analysis, and contextual labeling.
• HITL QA Engine: Human-in-the-loop workflows that route processed assets to human verification nodes to ensure scope-of-work compliance.
• Data & Vector Architecture: Petabyte-scale data ingestion using object storage, PostgreSQL for metadata, and Qdrant for semantic indexing and asset retrieval.

• The Robotics Frontier: Building data pipelines for embodied AI using real-world first-person environmental and motion data to train advanced robot policies.
• High-Throughput Refactoring: Converting experimental data science workflows into scalable, containerized production systems.
• Absolute Traceability: Maintaining end-to-end data provenance and integrity from asset capture through delivery to AI models.

• A short application read by the founders. We open your handles first: your GitHub with real commits, individual achievements, repos of projects you built.
• One conversation, then a build task from the real work.
• We watch how you approach the problem, how you break it down, what you measure, and how you verify your own work.
• Final conversation, then answer within 48 hours.