Dojo (Paymentsense)
Paddington, London
I spent my year in industry at Dojo, a UK fintech company that provides card payment technology to small and medium-sized businesses, working within Horizons, Dojo's in-house AI team, as part of a four-person engineering group that applies AI specifically to customer support operations. Over the placement I moved from a small, self-contained improvement on an existing internal tool to owning substantial pieces of a live, production AI system used daily by customer support staff, alongside contributing to how the team measures and improves the reliability of its AI outputs. Working in this small team using agile methods gave me early exposure to professional software delivery, from sprint planning and code review to translating business needs from non-technical stakeholders into concrete technical solutions. Beyond the specific projects, this placement reshaped my sense of what kind of engineer I want to be: I arrived expecting to pursue hardware engineering roles, but a year spent building and improving real AI products has shown me that software and AI engineering, particularly within fintech, is the direction I want to take after graduating. It has also been the clearest lesson of my degree so far that good engineering depends as much on clear communication with the people a system is built for as it does on the code itself.
I worked as a placement engineer within Horizons, Dojo's in-house AI team, as part of a four-person engineering group applying AI specifically to customer support operations. When I joined, my initial contributions were to Athena, a retrieval-augmented search tool that gives customer support advisors instant access to internal knowledge articles during live calls; the tool was nearly complete, and I worked on several improvements to it, including building a permissions system controlling who could edit handbooks, access full admin rights, or view a separate call analytics tool. My main project since then has been working on an AI-powered system that analyses customer support call recordings and automatically extracts useful information, such as what the customer needed and how the interaction went, into a searchable database accessible via AI Chat, which is now live and used daily by the support team. I also paired with a teammate to diagnose and fix a data-loss bug in the pipeline that measures how accurately the AI is performing, which had been silently dropping data above a certain volume. More recently, I've started work on a new tool that will give customer support advisors and managers AI-assessed feedback on call performance as well as complete compliance checks.
I learned Java and the Akka framework from scratch, which underpins the backend of Dojo's AI products, and gained hands-on experience with Google Cloud Platform services including BigQuery, Cloud Storage, and Datastore, which together store and organise the enriched call data. I worked with LangSmith, a tool for evaluating the accuracy of AI systems, including diagnosing and redesigning part of a data pipeline to fix a reliability bug that had been silently dropping data once export volumes passed a certain size. Overall, I gained a practical understanding of how large language models are integrated into real, production-facing products rather than just used in isolation.
This placement significantly developed my communication skills, particularly translating technical concepts for non-technical stakeholders in the customer support team so they could meaningfully shape decisions about tools that affected their daily work. Visiting the customer support team's office in Bristol twice during the placement helped ground this: seeing first-hand how advisors used our tools day to day made it much easier to understand their real needs and communicate effectively about trade-offs. Working within a small, four-person engineering team in an agile environment taught me how to scope and write clear tickets, manage my own priorities within weekly sprints, and balance independent ownership of tasks with close collaboration on shared, more complex features. I also developed more confidence and initiative in resolving problems independently, and a much clearer sense of how to identify a problem, propose an approach, and see it through to a working solution.