• Prompt and schema engineering against frontier multimodal models - getting reliable structured output from messy real-world data, diagnosing failure modes, and iterating on prompts that survive model upgrades.
• Agentic systems - designing and building agents that plan, use tools, and reason over our data.
• Realtime model optimization - working with high performance efficient models for edge analytics, filtering, understanding and event extraction
• Retrieval and semantic search - building the indexing, embeddings, and retrieval layers that make our RAG and agent systems actually useful, not just demo-grade.
• Computer vision - object detection, recognition, and other CV techniques applied to our video data, alongside (and sometimes instead of) multimodal LLMs.
• Statistical analysis and insight generation - turning the structured output our pipelines produce into trends, summaries, and product-surfaced insights. Picking the right tool for the question: a query, a classical model, an LLM, or a hybrid.
• Evaluation - labelled datasets, judge setups, regression tracking. A lot of the work is making "is this actually good?" answerable.
• The backend and infrastructure that holds it together - building the cloud services that run the pipeline reliably at scale.
• Audio and video understanding - transcription, speaker work, event detection, and the narrative/structured outputs the product surfaces.
• Some lower-level work in C/C++ and firmware where models need to run on-device under real-time constraints.

The mix shifts week to week. We're small, so you'll move between prompt iteration, backend code, eval tooling, retrieval/agent work, and looking at real data to understand why something went wrong.

What we’re looking for

• A few years of commercial (or equivalent) experience shipping ML-driven features in production - cloud, embedded, or both.
• Our stack spans Go (backend and ML orchestration), Python (offline eval, analysis, prototyping), and some C/C++ for edge work. However, solid experience in any relevant language is fine - we don't require prior expertise directly in these.
• Hands-on experience with LLM APIs and prompt engineering - not just calling the API, but knowing why outputs degrade and how to measure it. Bonus if you've built agents, tool-use systems, or non-trivial retrieval pipelines.
• Comfortable with data analysis - looking at examples, building labelled sets, choosing the right method (statistical, ML, or LLM-based) for the question, and drawing conclusions you'd defend.
• Good instincts for picking the right tool - when an LLM is the answer, when a smaller model is, when it's a heuristic, when it's "we don't have enough data yet."
• A strong CS/STEM background and the kind of curiosity that makes you dig into the data when something looks off.
• Frontend experience (Vue/React) is a bonus but not required.

Responsibilities

• Ship features against tight milestones, end-to-end where needed.
• Own the quality of the AI output you put into production - including the evals that prove it.
• Help shape how the team works as we grow.

We don't expect you to have done all of this before. If you've done a chunk of it well and want to learn the rest, please apply.

We believe that diverse perspectives make better products, and we strongly encourage people from underrepresented backgrounds in tech to apply. If you’re unsure whether you meet every requirement, please still reach out - we’d love to hear from you.

Why Antare

Join Antare in London and contribute to a team that’s creating the next generation of innovative products. We offer a competitive salary, equity options, and healthcare, along with the opportunity to build a product that makes a real impact.

While we offer some flexibility, we’re primarily an in-person team (4 days per week) - at this stage of the company, we’ve found that collaboration and ideation happen faster and more naturally when we’re in the same room.