AI / ML Team Leader

TL;DR
We're hiring a Leader for our AI / ML / Data Science team (US, California Bay Area or Houston preferred) to build and ship production-grade ML for industrial time-series and multimodal data. You'll spend most of your time hands-on (about 70%) delivering end-to-end solutions with customers, and the rest leading and scaling the data science team (about 30%). Position reports to the CTO and will have two direct reports.
The Role
We're hiring a Leader for our Data Science (DS) team to deliver production-grade ML for industrial systems using time-series and multimodal data (sensor/SCADA/historian, events, maintenance logs, and images/video where relevant). This is a hands-on role where you'll spend most of your time building and shipping (roughly 70% IC), while also leading and growing the team (roughly 30% people leadership).
You'll work closely with customers and domain SMEs, ship models to production,and evolve our concept from “models” to an autonomous decisioning system: forecasting → detection/diagnosis → optimization → closed-loop actions (with safety + governance). A key part of the role is advancing our unique IP in closed-loop autonomous operations.
What You’ll Do
Own the end-to-end lifecycle: problem framing → data readiness → modeling → deployment → monitoring → iteration.
Define and execute roadmap areas like anomaly/event detection, asset/process health, root-cause support, optimization, and closed-loop decision support.
Build scalable foundations for baselines, drift detection, model observability, and incident response.
Partner with industrial customers and SMEs to translate real process constraints into ML/optimization/decisioning solutions.Drive unsupervised/self-supervised initiatives (representations, clustering, change-point detection, weak supervision, active learning).
Develop a practical Reinforcement Learning (RL)/decisioning strategy (offline/safe RL, constrained optimization, simulators/digital twins), with guarded rollout patterns.
Lead and mentor DS talent, set processes, frameworks and quality standards (design/code reviews, documentation, postmortems).
Build and deploy AI / ML solutions / models in production.Own deployment, monitoring, performance validation, and iteration of models in production
Identify, document, and progress patentable innovations tied to closed-loop autonomy and production deployment.
Qualifications
8+ years in applied AI / ML technologies, including 2+ years leading teams (hiring, mentorship, performance management).
Deep experience with time-series ML at scale, ideally with messy industrial data [Ex: Frequency-domain time-series techniques (FFT/spectral analysis) and control/optimization methods (MPC-like approaches)].
Proven track record of shipping and operating AI / ML solutions in production (MLOps, monitoring, drift, retraining, reliability).
Strong Python and engineering fundamentals (clean code, testing, production patterns).
Strong communication, comfortable working directly with customers and cross-functional teams.
Bonus Points
Offline/safe RL, constrained optimization, and/or simulators/digital twins.
Self-supervised learning or foundation-model approaches for industrial time-series and multimodal fusion.
Robotics and / or Industrial domain experience (manufacturing, energy, chemicals, mining, utilities), including safety/uptime/latency/edge constraints.
Closed-loop or human-in-the-loop decision systems with governance and guardrails.
Experience contributing to IP strategy, invention disclosures, and patent filings.
Out of scope
Owning core product UI/UX design or front-end development.
Acting as the sole data engineer for ingestion/ETL across all customers (you'll partner closely with Engineering/Data Engineering where applicable).
Running IT/OT infrastructure, sensor hardware selection, or plant networking.
Doing research with no production path, success is measured in deployed outcomes and customer impact.
Being a full-time program manager, you will lead execution, but Delivery/CS/PM functions help run the overall program cadence.
What success looks like (first 6–12 months)
Customer impact: Delivered measurable improvements tied to customer KPIs (reduced unplanned downtime, improved yield, increased energy efficiency, lower maintenance cost).
Product impact: Delivered repeatable, documented DS approach, reusable feature/representation layers, evaluation harnesses, and monitoring dashboards.
Model quality: Improved anomaly precision/recall while reducing false positives, clear reduction in alert fatigue and better operator trust.
Time-to-production: Faster iteration cycles from prototype to deployed, monitored models, predictable delivery and fewer "one-off" solutions.
Production reliability: Strong monitoring, drift detection, and incident response practices in place, with clear ownership and runbooks.
Decisioning roadmap: A concrete, milestone-based plan for unsupervised/self-supervised and RL decisioning (simulation readiness, offline evaluation, gated rollout, safety/governance).
Team health: A high-performing DS team with clear standards, strong mentorship, and a hiring pipeline to scale responsibly.
About Kelvin
Kelvin.ai is paving a new path toward the future of industrial control applications. Our next-generation autonomous operations platform delivered as SaaS codifies the best engineering knowledge with the power of artificial intelligence. By combining human intelligence and machine learning, Kelvin delivers results across four pillars of customer value: Production Efficiency, Operational Efficiency and People Efficiency.

Employment type
Full-Time

Work arrangement
Yes