Data Pipelines & ML Engineer — Hybrid UK (EMI Options)

Theia Insights; Cambridge; Hybrid; GBP 90,000 - 140,000; Full time

Benefits offered by this job

EMI share options

25 days holiday + bank holidays

Private health insurance

Pension

Job summary

Theia Insights in the United Kingdom is seeking a data/ML engineer to build and operate production pipelines for global equity classification and knowledge graph work. You’ll run large-scale NLP/LLM inference, ingest market data, and shape schemas and data products used by downstream consumers.

You’ll collaborate with economists and engineers, deploy in a cloud environment, and apply embeddings, clustering and semantic similarity to production datasets, with a focus on reliability and cost

Qualifications

• Strong production Python experience.
• Pandas and Parquet/Arrow for datasets.
• Orchestrated batch pipelines (Dagster or Airflow) and S3-based data flows.
• Applied ML in production—embeddings and semantic similarity.
• AWS fluency and CI/CD discipline.

Responsibilities

• Build and maintain pipelines that classify global public equities across a five-level taxonomy.
• Run large-scale NLP and LLM inference over company documents with batch execution.
• Ingest market data and publish datasets to external distributors.
• Own schema and contract evolution for datasets with downstream consumers.
• Collaborate with economists and engineers to turn modelling decisions into production data.

Skills

Python

Pandas

Parquet/Arrow

DuckDB

Snowflake

Dagster

Airflow

AWS

CI/CD

ML in production

Embeddings

Tools

Snowflake

DuckDB

Job description

Theia Insights in the United Kingdom is seeking a data/ML engineer to build and operate production pipelines for global equity classification and knowledge graph work. You’ll run large-scale NLP/LLM inference, ingest market data, and shape schemas and data products used by downstream consumers.

You’ll collaborate with economists and engineers, deploy in a cloud environment, and apply embeddings, clustering and semantic similarity to production datasets, with a focus on reliability and cost