Artificial Intelligence Engineer

About the roleYou are a developer building internal tools for a trading business: the kind of tools that make smart people meaningfully more effective. Dashboards that surface the right signal. Automated alerts that catch what humans would miss. Lightweight apps that replace email threads and spreadsheets.The first chunk of work is real-time risk and trading tooling. Our trading-risk function (internally, TMT) does the live read on flow, positions and haircut through the day, and they are doing it without the visualisation and alerting layer they need. That is the immediate gap, and it is where you will make impact in the first few months.But the role is not bounded by one team. The same pattern - smart people, no tooling - runs through ops, the wider risk function, finance and the desks. As the early work proves out, your remit grows.This is not a core-system role. It is the support layer that sits next to the business and gives them eyes. We want speed, iteration and good visual judgement, not pristine architecture. If you are over-engineering this, you are doing it wrong.The wider opportunityYou will spend your time working alongside traders, risk, ops and the people who build our pricing and reference data. Most developers never get that kind of proximity to live markets.The domain knowledge you build here - how trading desks think, how risk is read, how flow moves, why a price is what it is - compounds across the rest of your career and is genuinely hard to acquire from the outside. As the platform matures, there is scope to move into deeper systems work, quant tooling, or markets-adjacent product if that is where you want to go.What you'll doBuild dashboards that surface the data the business actually needs - positions, haircut, VAR, exposure, P/L, utilisation - clearly, with the right level of zoom for whoever is looking.Build automated alerts and triggers: when a trader breaks pattern, when a position moves outside expectations, when a number that should reconcile doesn't.Replace email and Excel workflows with single submission and dashboard layers, so requests get filed properly and tracked.Work closely with whoever is pulling the thread at any given moment - initially the trading-risk function (TMT), then progressively across ops, the wider risk team, finance and the desks. They define the problem; you ship the tool.Iterate. A lot. The first version is wrong; the tenth version is what we wanted.Pick up the markets context as you go - how trades become risk, how risk becomes a number, how that number drives decisions. You will not be expected to know this on day one.Who you areYou ship fast. You are properly AI-fluent. Cursor, Claude Code or similar are your normal workflow, not a novelty. You use AI as a force multiplier without letting it slop out garbage.You have product instinct for data visualisation. You can tell at a glance whether a dashboard is usable or a wall of noise.You are comfortable in Python and a JS/TS stack (or equivalent). Rust is nice to have but not required for this role.You are curious about how markets work. You do not need to be a quant, but you should want to understand flow, risk, P/L and haircut, and you should enjoy talking to the people who live it.You are allergic to over-engineering when over-engineering is not the brief. You know when a notebook plus a Plotly chart beats a six-month build.You like working close to the business. This role does not work if you want to be left alone with a backlog.How you operateBiased to ship. You would rather have v1 in someone's hands today than v2 in your head tomorrow. Status meetings make you twitchy. You measure your week in things built, not tickets cleared.Low ego on your own code. You will throw away half of what you build because the user needed something slightly different from what you assumed. That does not bother you - you are attached to the outcome, not the artefact.Direct, both ways. You will tell a trader their idea won't work, and you will take being told the dashboard you spent two days on is wrong. Neither feels personal.Genuinely curious about the domain. You want to know why a price is what it is, why the trader does what they do, why the haircut matters. You are not a coder waiting for specs.Pragmatic. You know when good is good enough. You don't reach for a microservice when a Python script will do, and you don't ship spaghetti when something needs to last. You can tell the difference, and you don't sulk when told to do less.High-agency. You spot the problem before it is escalated. If a desk is doing something stupid in Excel and nobody has complained yet, you have already noticed and you are building.Matches our pace. Same hours, more output - that is the deal with AI now. We are impatient with slow. AI productivity flows to throughput, not slack. You are either energised by that or you are not.Nice to haveExperience in a commodities, energy, crypto or rates trading environment.Built internal trading, risk or ops tools before, even hacky ones.Familiarity with VAR models, Greeks, margin and haircut mechanics.Comfortable working off messy, evolving reference and pricing data.