Applied AI Engineer

You will work across the core systems that power our product: document intelligence, retrieval, agentic workflows, and the infrastructure required to deploy them reliably in production. This role is well-suited for someone who likes operating across layers, from messy PDF parsing problems to LLM workflow design, evals, and production deployment.
Responsibilities:
Document Intelligence: Build and improve the pipelines that turn complex financial documents into structured, usable data.
Retrieval and Agentic Workflows: Design and improve RAG systems, retrieval pipelines, and multi-step LLM workflows that extract, validate, reason over, and populate information into downstream outputs.
Evaluation and Reliability: Build the infrastructure to measure system quality. Define evals, failure taxonomies, and operational metrics that help us understand where workflows break.
Product and Infrastructure: Deploy scalable, resilient systems in production. Work closely with design, product, and business teams to rapidly build and iterate on user-facing workflows.
Requirements:
Strong programming ability in Python and TypeScript.
Experience integrating LLMs into production systems, including prompting, context management, structured outputs, and cost-performance tradeoffs.
Experience building or working with document processing systems such as VLMs for OCR and layout parsing models.
Comfort with cloud deployment and production systems, including containers, CI/CD, and Azure or GCP.
Experience thinking carefully about system quality, including evaluation, observability, or failure analysis for complex AI workflows.
Good to Have:
Experience with RAG systems, hybrid retrieval, reranking, and eval set design.
Experience with vision-language models or multimodal document understanding.
Familiarity with Azure or GCP-based AI infrastructure.
Experience building multi-step agentic systems or using modern agent tooling.
Genuine curiosity about financial services workflows, how investment banking, private equity, equity research, credit, or diligence actually work day to day.
Nice to Have:
You are AI-native. Tools like Claude Code, Cursor, Codex, and modern model APIs are part of your everyday workflow.
You know where they are powerful and where they fail, and you build with judgment around them.
You are an owner. Autonomous, self-directed, and comfortable with ambiguity. You take responsibility for outcomes, not just tasks.
You are energised by difficult problems, things that are technically hard, operationally messy, and valuable when solved well.
You want to understand how your users actually work. What makes a workflow painful, what accuracy really means in context, and why a product decision matters.

Experience
2-4 yrs