About the role
Responsibilities Design the reasoning layer: model context, tool calling and multi-step decomposition Build eval sets and harnesses, and analyse whether changes improve answers Choose models per task and manage token costs and caching Work with data engineers on semantic and signal layers Build and trace the agents and MCP servers behind the assistant Apply classical modelling such as anomaly detection and classifiers where it fits better than an LLM Requirements Shipped LLM features used in production Experience designing models with tools or structured data Built evals and analysed their results Strong Python with tested, modular services Strong SQL CI/CD, Docker and cloud experience, ideally Azure Tracing for LLM systems Quantitative reasoning with numbers Preferred Qualifications Fintech, lending or accounting background Spark or Databricks experience Semantic layers or metric views Fine-tuning experience Benefits Offers equity Hybrid or fully remote working with no return-to-office mandate About the Company Codat is an advisory intelligence platform for commercial banking, backed by JPMorgan, PayPal, Amex, Plaid and Shopify. It has powered over 350,000 connections to business customers' financial systems.