Founding Machine Learning Engineer

Stealth Startup; San Francisco (CA); On-site; USD 180,000 - 260,000; Full time

Job summary

Stealth Startup is hiring a Founding Machine Learning Engineer to design, train, deploy, and monitor end-to-end ML systems that fuse LLM-powered agents with time-series models. You will shape how agents interact with multimodal data and operate with speed and autonomy in a fast-paced startup.

You will champion production-grade ML, rigorous evaluation, and scalable infrastructure, collaborating with researchers, engineers, and key customers to drive rapid iteration and impact.

Qualifications

• 4–10 years of industry experience shipping production ML systems.
• Hands-on with LLM-powered agents, multi‑step reasoning, tool use, and autonomous workflows.
• Familiar with agent evaluation for reliability, safety, and metrics.

Responsibilities

• Design, train, and deploy production ML systems (LLM-powered agents + time-series models).
• Build and scale agents with multi-step reasoning, tool integration, autonomous workflows, memory, and adaptive strategies.
• Develop evaluation frameworks to ensure reliability, safety, and measurable performance.
• Apply time-series modeling techniques (forecasting, anomaly detection, multimodal data) in real-world scenarios.
• Operate end-to-end from data ingestion to deployment, monitoring, and continuous improvement.
• Stay up-to-date with AI agents, orchestration frameworks, and infrastructure (MCP, A2A).
• Collaborate with researchers, engineers, and lighthouse customers to validate solutions.

Skills

Founding ML Engineer

Production ML systems

LLM-powered agents

Agent evaluation

CI/CD

Observability

Reproducibility

Startup environment

Tools

MCP

A2A

CI/CD pipelines

Job description

We're hiring our Founding Machine Learning Engineer (MLE) with expertise in Agent Development and Time-Series Modeling. You'll play a foundational role in building production-grade systems that combine the power of LLM-powered agents with time-series foundation models.

The Role

This is not a narrow research role — you'll design, train, deploy, and monitor ML systems end-to-end, moving from prototype to production with speed and autonomy. You'll also be a core contributor to defining how agents interact with multimodal numerical data, a problem space where the playbook does not yet exist.

Job Description:

• Design, train, and deploy production ML systems (LLM-powered agents + time-series models)
• Build and scale LLM-powered agents with advanced capabilities: multi-step reasoning, tool integration, autonomous workflows, memory/context management, and adaptive strategies
• Develop and refine evaluation frameworks for agents, ensuring reliability, safety, and measurable performance
• Apply and extend time-series modeling techniques (forecasting, anomaly detection, multimodal fusion) in real-world customer scenarios
• Operate end-to-end: from data ingestion and preprocessing to deployment, monitoring, and continuous improvement
• Stay ahead of the curve on the latest innovations in AI agents, orchestration frameworks, and infrastructure (MCP, A2A, etc.)
• Partner directly with researchers, engineers, and lighthouse customers to validate solutions and drive rapid iteration

What we're looking for:

• Proven industry experience (4-10 years) as an ML Engineer, Research Engineer, or Applied Scientist, with a track record of shipping production ML systems
• Hands‑on expertise in LLM-powered agents: multi‑step reasoning, tool use, context windows, autonomous workflows, agent memory
• Deep understanding of agent evaluation techniques (reliability, safety, success metrics)
• Up‑to‑date with modern agent infrastructure and frameworks (MCP, A2A, etc.)
• Fluency with ML engineering best practices: reproducibility, monitoring, scaling, CI/CD, observability
• Comfort operating in a fast‑paced startup: shipping quickly, making tradeoffs, and thriving in ambiguity

Nice to have:

• Experience training custom neural networks beyond pre‑trained LLMs (e.g., transformers for time‑series or multimodal data)
• A background in time‑series modeling (forecasting, anomaly detection, classical + deep learning approaches)
• Published research or open‑source contributions in ML/AI