Principal AI Engineer

Our client is a leading global investment management firm headquartered in London, managing over $228B in assets. Technology, data science, machine learning, and AI are at the heart of its investment and research ecosystem.

The project focuses on building two key capabilities for secure and scalable AI adoption: Agentic Security and AI-Ready Data Foundations .

This is a hands-on engineering role where you will build AI agent workflows, data ingestion pipelines, evaluation frameworks, and guardrails that make AI outputs accurate, traceable, and reliable for investment professionals.

You will work with both structured and unstructured financial research data, identify where AI quality breaks down, and turn those insights into practical engineering improvements.

Most importantly, this is a greenfield initiative- the tooling does not exist yet, so you will have the opportunity to design and build it from the ground up.

Requirements

• Proven experience

building production agentic and LLM systems-

multi-agent or orchestrated workflows that reason across heterogeneous sources (PDFs, audio transcripts, file shares, databases) and surface confidence, gaps and provenance back to end users.

• Hands-on experience

engineering document ingestion and extraction pipelines

: parsing, chunking and the automated quality controls around them- detecting empty or truncated content, vendor feeds delivering the wrong section of a document, duplication, encoding and OCR defects.

• Experience

building evaluation and guardrail infrastructure

for AI systems: groundedness scoring, citation and provenance (file name plus the exact snippet retrieved), eval harnesses, regression suites and LLM observability.

• Strong production Python engineering-

services and pipelines that run unattended, with testing, CI and code standards. This is not a notebook-and-analysis role.

• Able to work from a deliberately vague brief, shape the problem directly with business stakeholders, and explain technical results to non-technical audiences.

Nice to have

• Experience tuning retrieval quality- chunking strategy, embedding choice, retrieval evaluation.
• Structured and time-series data-quality experience (coverage gaps, nulls in critical columns).
• ETL pipelines and fluency in SQL.
• Previous experience working with investment professionals in a fast-paced environment.
• Working knowledge of Snowflake, Linux/UNIX, Git, Jira.

Responsibilities

• *

Build agentic workflows

• that reason over research reports, transcripts, filings and news, and present portfolio managers with a clear view of what was found, what is missing, and how confident the system is in each answer.
• *

Engineer automated quality checks on unstructured source content

• before ingestion -empty or blank content, truncation, extraction fidelity, coverage gaps across expected document sets.
• *

Build vendor delivery validation

*: detect and quantify parsing and format defects in incoming feeds, feed them back to vendors and the data sourcing team, and fix extraction where it sits with us.

• *

Build evaluation harnesses, benchmarks and guardrails

• for agent output- groundedness, factual accuracy, relevance, and citation/provenance, so any claim can be traced back to a specific file and snippet.
• *

Ship monitoring and dashboards

• surfacing data-quality findings, confidence levels and coverage gaps to both engineering and PM audiences.
• Work directly with the platform engineering team, data sourcing, and portfolio managers to turn business expectations into measurable, automated quality standards.