About the Role
We are looking for a Senior AI Engineer to help evolve our existing AI platform for steel sourcing and quotation workflows. Our platform processes emails and documents, extracts structured product specifications, and uses AI agents to match customer requirements with supplier offers and available inventory. You'll improve these capabilities by combining steel domain knowledge, retrieval, and reasoning to produce accurate, explainable results.
Responsibilities
•Own and improve AI pipelines that extract and validate information from customer requests, supplier inventories, emails, PDFs, spreadsheets, and images.
•Build agentic matching workflows that interpret customer requirements, retrieve candidate products, compare specifications and constraints, and rank suitable offers - with clear explanations and human review where needed.
•Design and improve agent harnesses, including context management, tool access, workflow state, retries, stopping conditions, and human review.
•Build and operate Model Context Protocol (MCP) servers and clients that connect agents with tools, business systems, and domain knowledge, with appropriate authentication, permissions, monitoring, and error handling.
•Strengthen retrieval so agents interpret steel terminology and specifications correctly.
•Evaluate and integrate commercial and open-source models based on accuracy, latency, cost, and deployment requirements.
•Use Langfuse to manage prompts, trace agent behavior, and run evaluations that measure extraction and matching quality.
•Turn production failures and user corrections into evaluation cases and improvements to prompts, tools, and execution logic.
•Extend our Python services and integrate AI capabilities into existing product workflows.
•Work with business colleagues to prioritize problems, guide technical decisions, and mentor engineers while remaining closely involved in implementation.
Requirements
•Strong Python and backend engineering skills, with experience maintaining and evolving production systems.
•Hands-on experience shipping LLM applications and improving them after launch.
•Hands-on experience with all three: Pydantic AI or similar platform, LiteLLM, and Langfuse, covering agent orchestration, model routing and gateway operations, prompt management, tracing, and evaluations.
•Production experience building and operating MCP servers or clients, including authentication, permissions, monitoring, and error handling.
•Experience building reliable agent execution systems with clear tool interfaces, context and state management, failure recovery, and quality controls.
•Practical understanding of structured extraction, retrieval, tool calling, and AI evaluation.
•Practical experience building search, entity resolution, or domain knowledge systems.
•Experience with SQL databases, asynchronous processing, APIs, automated testing, and cloud infrastructure.
•A systematic approach to debugging across data, prompts, models, tools, and application code.
•The ability to make focused improvements in an existing codebase, explain architectural tradeoffs, and support other engineers.
Nice to have
•Agentic matching, recommendation systems, or ranking workflows that combine structured constraints, semantic retrieval, and domain knowledge.
•Deploying, serving, and optimizing open-source language models.
•Document processing, vision models, or multilingual extraction.
•Procurement, marketplaces, or industrial workflows.
•Prior steel industry experience isn't required. Curiosity and attention to detail are valued.
Conditions
You'll work on an existing product with concrete problems to solve and direct feedback on your work. You'll have scope to influence technical direction, improve how the team builds AI systems, and see your changes support everyday commercial workflows.
