Location: London (Hybrid), Safi HQ in Spitalfields Salary: £40,000 – £80,000 + equity Travel: Regular site visits across Europe, North America and Latin America Safi's mission is to make circular economy firms more profitable through the deployment of AI technology. We do that by developing foundational AI models, software, and data connectors. Role As an AI Product Manager for Vision Models, you are the critical layer that connects on-the-ground AI performance with our AI/ML engineering. You will: Capture and triage data + feedback from customer deployments Use your judgement and understanding of our customer's industrial processes + priorities, recycled material, and our vision AI capabilities to prepare and prioritise work for our AI/ML engineers Lead the domain discovery required to bring new materials and markets online Own the data pipeline and labelling teams that we use to teach our models You will spend time at recycling sites watching how operators actually assess quality - and translate what you see into precise AI model development, labelling and training sets. You must be comfortable planning with ML engineers, reading model-performance data, and reasoning about trade-offs with people. We think much of this role can be learnt on the job - high energy, commitment, quick thinking and willingness to get your hands dirty (quite literally, we work with waste materials) will go far. This is a junior-mid level role for someone who is excited about deploying AI that impacts real-life industrial plants. You do not need prior experience as a product manager in a tech company. Why Safi? Our customers are industrial recyclers of plastic and metals — manufacturers, processors, smelters. These firms are held back by limited, legacy technology and fragmented data. Within our first year of operations, we have signed customers such as one of the world's largest recycling plants, a top 5 global aluminium smelter and a group that processes the entire plastic waste stream of a major nation. We have product-market fit and strong customer traction in an under-served market. We want to expand on that traction to help manage the entire end-to-end lifecycle of plants in multiple sectors. We're backed by leading climate-focused VCs, including LowerCarbon Capital, Nosara Capital, and Transition Ventures. If our mission resonates with you, we encourage you to apply, even if your experience doesn't match every requirement. What You Will Do Own the ML backlog and prioritisation. Triage incoming data and feedback from customers and colleagues. Assess with sales, deployment and customers to understand importance. Execute the work - training preparation, data collection, labelling - that our ML engineers need to solve the customer problems Review model feedback and spot trends. Regularly review incoming AI-model feedback, identify recurring issues or trends, and feed the significant ones into the ML triage process. Maintain and curate the labelling guidelines (taxonomy). Keep the guidelines accurate and up to date as understanding improves and new edge cases emerge. Decide when to split out a new class and when to fold classes together — decisions that directly determine final model quality. Lead new-material and new-partner model development. Understand deeply with our customers how different objects, contamination and material affects their process. Understand what matters when assessing its quality: what can be judged visually, how accurate detection needs to be, and where the hard cases are. Gather detailed specifications, convey them to the ML team, sanity-check feasibility, create the labelling guidelines, and get annotators onto the work. Run weekly reviews with partners. Act as the interface between the business and the ML team. Run the weekly cadence with ML engineers, deployment, and sales so that priorities, guidelines, and progress stay aligned. What We're Looking For Required ML literacy: you have a basic understanding of how compute

AI Product Manager (Vision Models)

Don't apply unprepared