Senior Machine Learning Engineer
What we offer: An immediate opportunity to make an impact fighting climate change with a mission-driven team.
An in-person, collaborative culture. In our midtown Manhattan office, we not only have a stocked pantry but we also dedicate time to connect with each other during weekly happy hours and quarterly offsites.
National subsidized healthcare plans for medical, dental, and vision insurance.
Additional benefits include a 401(k) program, 12 weeks paid parental leave, and paid time off.
Free mental health and professional coaching appointments through Lyra .
At our ground-floor stage, our compensation structure places a strong emphasis on the value of high equity, with an annual compensation ranging from $150,000-$180,000.
What you'll do: Own, extend, and improve Thalo’s issue-detection engine spanning the electrical, refrigerant, and equipment-performance diagnostics at the core of our product
Research, develop, and implement ML, statistical, and LLM-based models in production, working directly with first-of-its-kind streaming sensor time-series data
Own our AI-evaluation practice: build labeled fault sets (from service outcomes, physics-vs-LLM disagreements, and field cross-checks), define accuracy metrics, and stand up an eval harness that regression-tests every prompt change, new detector, and model upgrade before it ships
Turn model outputs into clear, actionable insights and reports our field, CS, and BD teams can confidently put in front of customers
Continuously improve the data pipeline for large-scale ingestion, storage, transformation, and analysis so detection runs reliably and cost effectively as we scale
Partner closely with hardware, software, and business teams to connect field and customer insights back into the product and document your work so the whole team can build on it
What you have: 5+ years building and deploying ML or statistical models on production data, ideally in an early-stage startup environment
M.S. or higher in a quantitative discipline such as math, physics, statistics, or data science (or equivalent applied experience)
Strong applied experience with time-series or streaming sensor data, including anomaly detection, forecasting, signal processing, or similar
Hands-on experience shipping production features on frontier LLMs (e.g., prompt engineering, structured output, tool-use/agents, and RAG) with the judgment to know when an LLM is the right tool versus a deterministic rule or a statistical model
Experience evaluating AI systems: building eval sets, measuring precision/recall, using LLM-as-judge, and guarding against regressions as prompts and models change
Fluency in Python and the modern data stack, with the software-engineering chops to ship production-grade code (not just notebooks)
A real customer instinct: the ability to translate a model output into a plain-English insight a technician or building operator will trust and act on
Curiosity about the physical world and the drive to understand the “why” behind the product, not just how to implement it
A self-directed, ownership mindset and a habit of documenting and sharing context
Bonus points: A passion for tackling climate change and promoting sustainability
HVAC, refrigeration, combustion, building-systems, or energy-domain experience (a strong plus, but something we’re happy to help the right person learn)
Experience with agentic / tool-use systems, RAG over technical documentation, or LLM vision
Familiarity with LLM cost/latency optimization (prompt caching, batch inference) and model governance (managing upgrades, monitoring output/score drift, A/B-testing context changes)
Frontier-class LLM, open source LLM, and/or AWS Bedrock in production
Full-stack comfort to take a feature to the UI (React/TypeScript); time-series databases (InfluxDB, TimescaleDB) and tools like Grafana; a degree in a quantitative or engineering discipline