Staff AI Engineer

What you will own

• Own the AI and intelligence layer and the engineering discipline that keeps it grounded and honest.
• Own the machine-learning models that learn each customer’s patterns.
• Own LLM and agentic systems that produce grounded natural-language output.
• Build per-customer models without leaking the signals they are meant to detect.
• Set a high bar for evaluation and grounding across the layer.
• Set technical direction for the AI layer.
• Partner closely with the evaluation engineer to make quality measurable.
• Turn user feedback into durable improvements over time.

What you bring

• Staff-level applied ML/AI experience in production.
• Strong classical machine learning with hands-on depth in time-series modeling, anomaly and outlier detection, and gradient-boosted models such as LightGBM or XGBoost using Python plus scikit-learn and pandas or Polars.
• Modern LLM and agent patterns, including retrieval-augmented generation (RAG), tool-using agents, schema-validated structured output, and orchestration with LangGraph or a similar framework.
• Rigor on evaluation, grounding, and hallucination control, including measuring whether an LLM’s output is faithful to source data and building guardrails that prevent unsupported statements from reaching users.
• Discipline around data leakage and honest evaluation, including designing features, splits, and validation so models are never trained or scored on information unavailable in real use.
• Clear thinking about what ML outputs can and cannot claim, including explaining to non-ML colleagues where evidence ends and signals begin.
• Real software engineering, including writing tested, production-grade Python services, comfort with SQL against warehouses such as Snowflake or PostgreSQL, and shipping containerized software.
• Technical leadership, including mentoring engineers and working across backend, data, product, and evaluation colleagues to convert feedback into lasting improvements.

Key technologies

• Python, scikit-learn, pandas, Polars, LightGBM, XGBoost
• LangGraph, RAG, LangChain
• SQL, Snowflake, PostgreSQL

What makes this role different

• Grounding is a hard requirement: no fabricated or unsupported output, ever, with faithfulness treated as a gate.
• Separation of concerns across ML surfaces: generated text only states what is supported by verified data or confirmed by a human.
• Per-customer intelligence models that learn business behavior with strict anti-leakage discipline.
• Customer-cloud deployment, including managed LLMs in-VPC, so customer data never leaves their environment.

Nice-to-haves

• Finance or accounting domain experience.
• Feature-attribution literacy, including experience with SHAP.
• Human-in-the-loop systems, including designing review, confirmation, or correction workflows and feeding them back into models.
• Managed LLMs inside customer clouds, including experience running AWS Bedrock, Google Vertex AI, or Azure OpenAI inside a customer’s VPC.
• Sparse and zero-inflated data.
• Multi-deployment enterprise software experience across multi-tenant SaaS, single-tenant SaaS, and on-premises deployments across AWS, Azure, and GCP.