Staff Engineer- Applied AI

About the Role

AI is not a feature in Foundry - it is the operating model. AI-assisted coding and product management is baked into how the team works, and AI-assisted solutioning is baked into what we ship: LLM-based data ingestion that auto-fills the Sizer from unstructured customer files, a constraint-solver recommendation engine that suggests robot configurations, semantic search over historical sizing sheets, and a Copilot that guides users through the solutioning journey. As the Staff Engineer, Applied AI you own all of this. You run the Applied AI capability as a chapter across all Foundry squads- setting the patterns, evaluation frameworks and API contracts that each squad implements rather than centralising all AI work in one team. You are expected to be the most technically credible AI engineer in the organisation, influencing architecture decisions across squads and setting the standard for how AI features are built, measured and shipped.

WHAT YOU'LL DO

• Build and own the recommendation engine - Design and implement the ML pipeline that recommends solution type (technology, layout, configuration) from customer inputs; own the feature store, training pipeline, model registry and serving layer.
• Build the constraint-solver intelligence - Implement the AI layer on top of the constraint solver: learned priors for solver warm-starts, anomaly detection for unusual input combinations, and automated sensitivity analysis.
• Build LLM-based data ingestion- Design the production pipeline that takes unstructured customer documents (PDFs, spreadsheets, emails) and extracts structured Sizer inputs using LLMs; handle uncertainty, partial extractions and human-in-the-loop review.
• Build the formulator engine and vector DB- Implement semantic search over historical sizing solutions and L1 data sheets; design the embedding pipeline, vector database and retrieval-augmented generation (RAG) layer.
• Define AI chapter patterns- Set the shared patterns that all squads adopt: LLM API integration (streaming, function calling, structured output), prompt engineering standards, evaluation harnesses and A/B experiment frameworks for AI features.
• Build the AI evaluation framework- Implement offline and online evaluation for every AI feature: golden-set benchmarks, production-shadow evaluation, A/B experiment readout tooling and regression detection.
• Own the AI API contracts- Design the internal API contracts between the AI squad and consuming squads (Sizer, Layout, GCM); version them, document them and enforce them via contract tests.
• Mentor and grow AI capability- Pair with and review engineers across squads who are implementing AI features; run the Applied AI chapter meetings; grow the team's collective AI engineering capability.

MINIMUM (REQUIRED) QUALIFICATIONS

• 10–15 years professional engineering; at least 4 years in applied ML / LLM productionisation.
• Deep experience with LLM APIs (OpenAI, Anthropic, or open-source): function calling, structured output, streaming, embeddings.
• Production RAG system experience: chunking strategies, embedding models, vector databases (Pinecone, Weaviate, Qdrant, pgvector), retrieval evaluation.
• Experience building and deploying recommendation systems or ranking models.
• Strong software engineering foundations: API design, observability, testing, CI/CD.
• Track record of cross-team technical influence: setting patterns others adopt.
• Experience with ML experiment tracking and model-lifecycle management.

PREFERRED QUALIFICATIONS

• Experience with optimisation / operations-research methods used in conjunction with ML (hybrid AI + OR solvers).
• Experience with agentic AI systems: tool-use, ReAct, AutoGen, LangGraph or similar.
• Domain experience in warehousing, logistics or industrial automation.
• Experience with online A/B experimentation for AI features (including metric design).
• Familiarity with structured-extraction frameworks (Instructor, Outlines, Marvin).
• Experience with feature stores (Feast, Tecton) and ML platforms (MLflow, Vertex AI).
• Prior Staff-level or principal-equivalent scope at a technology company.

TECHNICAL QUALIFICATIONS

We expect you to be proficient in or quickly learn the following. Senior roles (Band E+) are expected to be experts; Band C/D roles are expected to have working knowledge.

• LLM stack: OpenAI / Anthropic API; LangChain / LlamaIndex / LangGraph for orchestration; structured output parsing (Instructor / Pydantic).
• RAG: chunking and indexing pipelines; embedding models (text-embedding-3, Cohere, BGE); vector stores (pgvector, Qdrant, Pinecone); hybrid search (dense + BM25).
• ML: scikit-learn, XGBoost for classical recommendation; PyTorch / HuggingFace for fine-tuning and inference; MLflow or Vertex AI for experiment tracking.
• Evaluation: RAGAS, TruLens, or custom harnesses; LLM-as-judge pipelines; A/B frameworks integrated with analytics (e.g. Mixpanel).
• Languages: Python (primary); Java or Go for production service wrappers.
• APIs: REST (FastAPI or Flask); gRPC for high-throughput internal calls; SSE / WebSocket for streamed LLM responses.
• Infrastructure: GCP (Vertex AI, Cloud Run, GCS, Pub/Sub); Docker, Kubernetes; Terraform.
• Observability: LLM call logging (token counts, latency, cost); Prometheus metrics; Grafana dashboards; structured logging with trace IDs.
• Testing: unit tests for extraction logic; golden-set regression tests; integration tests with mocked LLM responses.
• Tooling: GitHub monorepo, GitHub Actions CI, Backstage.

HOW WE WORK

You are the single senior AI voice across Foundry. You lead the Applied AI chapter- short, weekly craft syncs with engineers across squads who are implementing AI features and you own the AI API contracts that every squad consumes. You report directly to the Senior Director and participate in the weekly squad-of-squads sync. You are expected to influence decisions at the architecture level, not just the implementation level.

Department

GreyMatter Engineering

Mandatory Skills

Python,Machine Learning,LLM,AI

Experience

8 to 15 years