Founding ML Engineer

a16z-speedrun•San Francisco, CA
•$130,000 - $150,000

About The Position

Philyron is building a predictive platform for commercial fleets, including tankers, cargo ships, and defense vessels. It uses real-time sensor data and historical maintenance records to predict and explain engine failures before they happen. The Founding ML Engineer will own the AI systems at the heart of the product and work directly with the founders. The role spans the full machine-learning lifecycle, including data, model training, evaluation, and production deployment. Responsibilities include building reliable data pipelines from vessels to the cloud over intermittent satellite connections; establishing AWS infrastructure for real-time alerts and time-series storage; integrating historical maintenance data from operators' CMMS databases; developing predictive models that combine live sensor signatures with past failure records; and creating scheduled reports for operators.

Requirements

  • Evidence of having shipped machine-learning work that real users have used, whether in production, research, or serious projects.
  • Demonstrable projects, repositories, demos, or side projects designed and built end to end, ideally involving both data or machine learning and supporting infrastructure.
  • Strong fundamentals across the stack, beyond notebook-based work.
  • Pragmatic approach to choosing reliable, straightforward solutions over unnecessarily clever ones.
  • Comfort owning systems end to end without an existing machine-learning team.

Responsibilities

  • Build reliable pipelines to transfer sensor data from vessel edge devices to the cloud over satellite connections such as VSAT and Starlink, handling intermittent connectivity and resumable transfers.
  • Use lightweight, fault-tolerant protocols such as MQTT and Sparkplug B, with encryption in transit.
  • Stand up AWS cloud architecture, including a real-time processing path for checking sensor metrics against safety thresholds and sending alerts.
  • Build a storage path for high-frequency time-series data used for trend analysis and model training.
  • Develop an integration layer to retrieve historical maintenance records from operators' CMMS databases.
  • Build machine-learning systems that combine live sensor signatures with historical failure records.
  • Map past maintenance events to preceding sensor patterns and train anomaly-detection and failure-prediction models.
  • Design fleet-wide and per-vessel modeling approaches to produce accurate, engine-specific predictions.
  • Build a service that compiles trends, history, and predictions into scheduled reports for operators.
  • Own the end-to-end AI system, from data and training through evaluation and production deployment.
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service