Machine Learning Engineer

ABOUT DESCRIAL Descrial is an AI work-execution platform purpose-built for alternative investment firms and their portfolio companies and securely orchestrates proprietary data, external sources, and workflows. Descrial is an important pillar of Enterprise Intelligence Orchestration (EIO) and enables it by structuring and standardizing data while embedding domain expertise, thereby enabling faster and contextual analysis, broader coverage, and more intelligent decision-making. It is backed by TresVista's 1,800+ employees and deep domain expertise across investment diligence, research, valuation, fund administration, accounting, and data analytics. The team is small, senior, high-ownership, and in-office, with direct access to the Executive Committee. OVERVIEW As a Machine Learning Engineer on the UK core team, you will design, build, and evaluate the LLM-powered systems at the heart of W-OS: multi-step agentic workflows, retrieval over confidential investment corpora, document intelligence, and the evaluation and observability infrastructure that makes them trustworthy in production. You will work hands-on with LangGraph and LangChain across leading model providers, on a Python/FastAPI backend with vector and graph retrieval, where token cost, latency, and auditability are first-class engineering constraints. We are hiring four people at two levels — senior engineers owning major capabilities, and lead engineers who additionally set technical direction and mentor. ROLES AND RESPONSIBILITIES Agentic and LLM systems - Design and implement multi-step agent workflows in LangGraph / LangChain: state management, checkpoints, tool-calling, and human-in-the-loop patterns - Build structured-output and prompt systems (Pydantic-validated) for extraction, summarisation, classification, and reasoning over investment documents - Integrate model providers (Anthropic, OpenAI, Amazon Bedrock) with production discipline, and own model routing and fallback: choosing the right model per task, degrading gracefully when one is unavailable, and absorbing provider model changes without breaking users - Treat context as a managed resource: what enters a window, compaction across long agent runs, and controlling cost through context design as well as model choice - Build and consume MCP (Model Context Protocol) tool servers giving agents safe access to internal services, search, and document stores Retrieval and knowledge systems - Own RAG pipelines end to end: parsing, chunking, embeddings, vector retrieval (Qdrant), reranking, and hybrid search - Extend knowledge-graph-backed retrieval (GraphRAG) on Neo4j, including Cypher and graph data modelling - Tune retrieval quality with rigorous relevance evaluation and multi-tenant-safe isolation of indexed content Evaluation, quality, and cost - Design and run evaluations: golden datasets, LLM-as-judge patterns, and regression gates on prompt or model changes - Instrument every model call (LangSmith / LangFuse class tooling) for latency, cost, and quality in production — token-cost engineering is a core part of this job - Apply responsible-AI guardrails: tool-use constraints, output validation, and data-privacy awareness in a UK/GDPR context - Ship supporting services in Python 3.11+ / FastAPI with PostgreSQL persistence; operate on AWS through CI/CD with quality gates WHO YOU ARE Prerequisites - 5+ years of software engineering with strong Python, including 2+ years building LLM or ML systems that ran in production - Deep hands-on experience with an agent/orchestration framework (LangGraph, LangChain, or comparable) beyond prototypes - Production RAG experience: embeddings, vector databases, chunking strategy, retrieval evaluation - Graph-backed retrieval experience, or the graph data-modelling fundamentals to pick up Neo4j and Cypher quickly - Evidence of evaluation-led development: you can describe the eval harness of your last system from memory - You build with AI coding tools daily (Cursor, Claude Code or comparable) and have a considered view of where they help and where they do not Good to have - Qdrant, Weaviate, or pgvector in production; document-ingestion stacks (Docling, unstructured) - MCP server development; LLM observability tooling (LangSmith, LangFuse, Ragas) - Fine-tuning, embedding-model selection, or open-weight deployment experience - Financial-services document domains (filings, fund documents, research) EDUCATION Bachelor's degree in engineering, computer science, or a related field, or equivalent experience from a strong technical background. WHAT WE OFFER - You will be working closely with the Executive Committee who have over 20 years of experience in leading a company with 400+ leading clients in the alternative investment space - Competitive compensation - Opportunity to build something impactful for established and leading companies in the financial services industry - Innovative, supportive, and collaborative work environment - Office in central London (King's Cross) HOW WE HIRE - Stage 1: Introductory conversation - Stage 2: Technical assessment - Stage 3: Interview with a senior member of the team - Stage 4: Final interview with a member of the Executive Committee