Data Scientist & LLM Engineer

Turn millions of video events into insight, and build vision-language and LLM workflows on top of our media pipeline.

Our platform produces a rich stream of recordings, detections, tracks, events, and operational context. As a Data Scientist & LLM Engineer, you will turn that data into dependable features and build language- and vision-language workflows that help users understand what happened.

You will combine product thinking with careful experimentation. The goal is not a clever demo: it is an evaluated, observable system that behaves predictably with real customer data and makes its uncertainty visible.

Your impact

What you will work on

• Build vision-language and LLM workflows for summarisation, event understanding, search, and operational automation.
• Design evaluations, test sets, and feedback loops that measure quality, safety, latency, and cost.
• Explore event and media metadata to identify patterns that can become useful product capabilities.
• Develop retrieval and context strategies across structured data, model results, and selected media evidence.
• Productionise experiments as tested services or workflow stages with monitoring and clear failure behaviour.
• Work with product and engineering colleagues to scope problems, review output quality, and iterate from user feedback.

About you

You may thrive here if

• You are strong in Python and comfortable analysing data as well as building production software.
• You have practical experience with LLMs, vision-language models, retrieval, or adjacent applied-ML systems.
• You know how to evaluate probabilistic systems beyond a few hand-picked examples.
• You can work with structured and unstructured data and reason about provenance, access, and data quality.
• You communicate trade-offs clearly and prefer simple, testable approaches over unnecessary framework complexity.

You do not need to match every point. If the work sounds like a strong fit, tell us what you would bring and where you want to grow.

Useful additions

Helpful experience

• RAG, tool use, agent harnesses, or structured-output workflows
• Multimodal models that combine text, images, or video
• Media pipelines, computer vision, or time-series analytics

How we work

Skills differ by role. These are the behaviours we expect from everyone building UUG.AI.

• Strong communication

  Share context, decisions, and concerns clearly. Ask questions early, listen carefully, and adapt the message to the people involved.

• Disciplined and honest

  Do what you say, work with care, and be direct about uncertainty or mistakes. We value evidence and transparency over appearances.

• Team player and owner

  Help the team succeed while taking responsibility for the outcome. Collaborate openly, follow through, and leave the work better than you found it.

A clear start

What to expect in your first months

The exact pace depends on the role and your experience. We use these steps to align on support, ownership, and useful outcomes.

• Understand the data

  Map the event and media pipeline, review current AI workflows, and learn which outputs users need to trust.

• Establish an evaluation baseline

  Turn a product question into a representative test set and an explicit quality, latency, and cost baseline.

• Ship and observe a workflow

  Own a capability from experiment to production, including feedback collection and a plan for continued evaluation.