Data Scientist, AI Agents ( 60-100k + Equity) at Arctal
Data Scientist, AI Agents Salary: £60–100k + Equity Company Description: Arctal - AI startup building high-fidelity financial datasets Job Description: You will act as a super-IC managing a fleet of AI agents to transform 100,000+ unstructured financial PDFs into clean, queryable data for institutional buyers. This role sits at the intersection of data science and engineering, requiring you to design agentic workflows, prompt chains, and validation logic to ensure professional-grade accuracy at scale. Location: London, UK Why This Role Is Remarkable:
- Work at the frontier of agentic workflows, using tools like Claude Code and Cursor to automate the extraction of complex data that was never meant to be machine-readable.
- Join an elite team of five with the output of fifty, founded by industry veterans from Secondmind and Sylvera with deep roots in Cambridge ML research.
- Enjoy high-intensity growth and direct exposure to institutional customers like asset managers and banks, with meaningful equity in a fast-shipping startup.
What You Will Do:
- Design and maintain agent-led data pipelines using prompt chains and human-in-the-loop checkpoints to ensure 100% data reliability.
- Build validation logic in Python and SQL that encodes your expert judgement to catch the edge cases and hallucinations AI agents might miss.
- Debug and optimize complex data ingestion and transformation systems, acting as the final line of defense before data reaches global financial clients.
The Ideal Candidate:
- Possesses 1–5 years of experience as a data scientist or analytics engineer with a track record of building production-grade data pipelines, not just querying tables.
- Is an AI-native practitioner who uses agentic coding tools like Cursor or Claude Code for real-world engineering tasks rather than simple chat experiments.
- Demonstrates an obsessive attention to detail and technical depth in Python, SQL, and terminal environments, with a passion for defining "ground truth" in messy datasets.