Product Manager, Data Engine

2 нед. назад

USAMiddleOnsite

data analysisproduct roadmaptoolingplatform management

Responsible for owning the deployment platform including the data engine powering robot training and internal/external tooling for fleet management.

Обязанности

We're hiring a Product Manager, Data Engine to own our deployment platform. The core of it is the data engine: everything that turns what happens in the field into training signal our models can learn from. It is the highest-leverage system we have for how fast our robots improve.
The same role owns the tooling built around it, both the tools our field teams and customers use to run robots at their sites and the internal platform our research, engineering, and operations teams work in every day. At Dyna, internal tools are core products, and you'll own their roadmap and quality bar. Together, this is what turns working robots into a working fleet.

Будет плюсом

Experience labeling video, multimodal, or time-series data rather than only images or text.
Background in robotics, autonomous vehicles, drones, or other fleet operations, particularly labeling perception or behavior data.
Experience with human-in-the-loop systems, teleoperation, or remote assistance products.
Experience with deployed hardware in customer environments and the operational reality that comes with it.
Experience with multi-site robot deployments.
At , we build technology for the real world, which requires a team as diverse as the environments our robots inhabit. We are an equal opportunity employer committed to technical rigor and mutual respect.
Don’t let a checklist stop you. Data shows that underrepresented groups often only apply if they meet 100% of the criteria. We value problem-solving and grit over keyword matching. If you’re passionate about the intersection of data and robotics, we want to hear from you – even if you don't check every box.

Другое

builds general-purpose robots powered by a proprietary embodied AI foundation model with top-in-industry generalization and real-world performance. Already deployed with customers across multiple industries, our robots do commercial-grade work in the physical world. Our team comes from Google DeepMind, Meta, and Cruise, and we're backed by CRV, First Round, and other leading investors.
Decide what we capture. Define what every robot logs and what an episode has to contain for someone to diagnose a failure, then get it instrumented across the fleet.
Own the labeling schema. This is how we describe what robots do and what goes wrong, and it is where the job starts. A schema that matched the field six months ago won't match it today, and a pipeline running at high coverage against a stale schema is worse than useless because it looks healthy. You own the definitions, the audits, and the call on when it has to change.
Own the quality bar. Define how label quality is measured and audited, whether the work happens in-house or with vendors, and hold the line when coverage and quality pull against each other. That includes knowing where model-assisted labeling helps and where it quietly degrades quality while the coverage numbers look fine.
Own the rest of the loop. Episode and telemetry capture, annotation pipelines, model-behavior observability, and remote teleoperation for capturing recovery and intervention data. This is the loop that makes each deployment better than the last.
Own the tools our own teams live in. Labeling and visualization systems, model evaluation tooling, and task management, used daily by researchers, engineers, and operations.
Design for technical, opinionated users. They want speed and direct access rather than guardrails, and they'll route around anything that slows them down. Own the UX and not just the capability: find the pain points these teams have stopped complaining about, and make them measurably faster.
Own the tools that run robots on the floor. Setup, monitoring, evals, intervention, and escalation, used mid-shift by people with no technical background and no time to spare. That constraint shapes every decision in this area.
Own the customer-facing view. Decide what customers need to see to trust robots running in their business, which internal surfaces are on a path to them, and what has to change before they get there.
Spend real time on site. Watch robots fail and operators work. What a hotel GM needs in order to trust a robot on their floor is different from what our field team needs in order to fix one, and the requirements that matter most here are the ones nobody files a ticket about.
4+ years of product management experience owning labeling, annotation, or training data pipelines for ML systems end to end, from what gets captured through to what a model trains on.
Familiarity with ML infrastructure and how training pipelines consume the data you produce.
Experience designing and building tooling from scratch, both for technical users and for non-technical people doing time-sensitive work in real sites.
Comfort with systems that fail in physical ways. Robots break, sites are messy, and problems are rarely reproducible from a laptop.
Willingness to travel to customer sites regularly, and to be genuinely useful when you're there.