Voting Data Analyst - tumelo

Job Title : Voting Data Analyst

Start date : ASAP

Location: Bristol/Hybrid (At least 2 days a week in-office)

The Company

Tumelo is a fintech on a mission to empower every investor to shape a better world. We build the infrastructure that lets everyday investors — pension members, fund holders — have a real say over how their money is used at company AGMs, even when they invest through pooled funds.

Our latest product, ProxyBeacon, uses AI to generate bespoke proxy research and voting recommendations for asset managers and asset owners — replacing the one-size-fits-all approach of traditional proxy advisors with something faster, cheaper, and genuinely tailored to each client's priorities.

The Team

Tumelo is committed to building an inclusive, diverse and equitable culture based on the principles of high trust, autonomy, empowerment and psychological safety. We take inspiration from our company values - daring, nimble, mission-driven, empowered and supportive. We all strive to represent these values in the way we work with our colleagues and external stakeholders alike.

The Role

ProxyBeacon requires reliable, accurate data from company proxy filings in order to generate research reports and voting outcomes. The quality of the data — how well they're defined, how precisely the AI is instructed to find them — determines the quality of everything downstream.

That's where you come in.

You'll work in a team with responsibility for the data point layer: understanding, extracting and testing the data that our end users - stewardship specialists - need to carry out their duties. You’ll be writing prompts for LLMs, manually checking, and iterating. It is a role for someone who is curious, precise and takes real pride in getting things right.

What You'll Do

• Write and refine data point descriptions — clear, structured definitions that instruct the LLM what to extract from proxy filings and how to handle edge cases
• Review and improve prompts — sense-check that prompts accurately reflect the intended data point description and surface any misalignments
• Run test cases — use our tools skills to run extractions across a range of documents and manually evaluate the results
• Iterate — when results aren't right, work through why and fix it
• Ensure fitness for purpose — understand how each data point feeds into downstream policy rules, and flag where a definition needs to be tightened to make the rule work
• Liaise with engineers — communicate clearly with the engineering team when issues require changes to the underlying RAG process, translating between the technical and the business context

What You’ll Bring

We'll train you on the domain and the tools — what we can't train is the mindset.