Careers
Come play for real.
The athlete mindset.
Our culture
We look for athletes, whatever the sport: people with a relentless pursuit of growth and excellence. Here the scoreboard is a client's P&L, and the fans are the operators whose day gets better when the system works. Three things describe how we work: how we train, how we compete, and how we last.
How we train
Fundamentals, every day.
- The unglamorous work runs the business: data that reconciles, rules that hold, integrations that survive month-end. We are obsessed with getting it right.
- Small teams, short loops, weekly quality floors. We measure before we argue and ship before we polish.
- Everyone stays at the edge of the field: the latest models and agent tooling in daily use, brought back to the team.
How we compete
One team, play to win.
- We play to win: a number moving in the client's P&L, not an interesting pilot. The scoreboard is agreed before kick-off, and their numbers decide what scales.
- We win as one team with the client. We sit with their operators, design with the people who will use the system, and share the result, good or bad.
- No titles, no hand-offs inside the firm. The engineer who sat with the quoting desk fixes the quoting desk, and the best idea wins whoever brought it.
How we last
Built for a long season.
- Deployments are intense: weeks on site, real operations, real stakes. We protect the recovery between them, because tired teams make expensive mistakes.
- We stay past the month-six gate. Each system makes the next one faster, and compounding beats a quick win.
- Our fans are the operators: the buyer, the planner, the claims handler. We build with them and measure ourselves on what they get back.
Elite team. Real operations. Measured results.
What it is like here
A small, senior team that puts AI to work inside real operations: quoting desks, claims queues, warehouses and finance back offices. Every engineer owns a use case from the first conversation to the number it returns, and the client's own P&L is the review that counts.
Understand the problem first
Whatever your role, engineer, researcher or deployment lead, you start by understanding the client's business problem: how the work runs today, who does it, and which number it moves. Nothing gets built until that is clear.
Get close to the work
You will spend time in warehouses, planning rooms and back offices, not only in an IDE, alongside the people who do the job. The best systems come from understanding how the work actually runs.
Work with the frontier
We work with research labs at leading universities and use the latest models and agent tooling daily. You will be expected to stay at the edge and bring it back to the team.
Own the outcome
Every engineer owns a use case from scoping to the measured result. The baseline, the target and the number at the end are yours, not a ticket's.
Measure before you argue
Holdouts, evaluations and weekly quality floors. A client's own numbers decide what scales, and ours decide what we keep building.
Three offices, one team
San Francisco, Singapore and Shenzhen, with regular time at client sites. One team, no hand-offs: the people in the first meeting do the work.
Open roles.
Hiring now
Engineering, research and deployment. We do not believe in titles: every role works directly with clients and with the partners, and your scope grows with what you deliver. If you do not see your role but believe you belong here, apply anyway and tell us why.
Deployment Deployment Lead San Francisco · at client sites
You will own the non-code half of a deployment: discovery, process mapping, training, change management and adoption. You will partner with forward-deployed engineers, map how the operation actually runs, and make the change stick with the people doing the work.
What you will do
- Run discovery. Sit with client teams to understand their processes, pain and goals, and turn that into a clear plan for the scan and the pilot.
- Map how work flows. Document how quoting, service, procurement, finance or logistics actually runs so the system fits reality instead of fighting it.
- Drive adoption. Lead training, build the materials, and do the hands-on change management that gets operators to trust and use the system day after day.
- Own the relationship. Be the reliable point of contact through go-live, keeping stakeholders aligned and momentum high.
- Prove the outcome. Track the numbers that matter to the client and feed lessons back into the next deployment.
What you will bring
- 7+ years in strategy consulting, or in a deployment or forward-deployed role at a software company, including leading engagements end to end
- Genuine interest in traditional industries: distribution, manufacturing, logistics, services
- Highly organised; you keep a complex project with many stakeholders moving
- Build trust easily with both operators and senior executives, and care more about the change sticking than winning the room
- Bonus: familiarity with ERP-driven processes; Mandarin or German
Engineering Forward-Deployed Engineer San Francisco · Singapore · Shenzhen · at client sites
We are building an elite, talent-dense team that puts AI systems into the operations of real enterprises. We need end-to-end engineers who can go from a client's warehouse or quoting desk to shipping production code in the same week. You will embed with operators, turn what you learn into working systems, and own each use case from scoping through the numbers it returns.
What you will do
- Embed with clients. Sit with sales, service, procurement and finance teams to understand their hardest operating problems firsthand, on the floor and in the planning room.
- Ship into daily work. Take a use case from scan to pilot to production: connect the systems, model the objects, build the workflow, and land it in the team's actual tools.
- Own the outcome. Be accountable for the baseline, the target and the measured result, not just the code.
- Feed the model. Turn what you learn on each deployment into reusable objects, rules and evaluations so the next use case starts further ahead.
- Set the standard. Help define how applied AI gets built and adopted here, and what excellent looks like.
What you will bring
- 3+ years as a full-stack or backend-leaning engineer, with production code and real stakeholder outcomes behind you
- Experience building with LLMs and agent systems, not only using them
- Strong system design instincts, and daily use of AI coding tools
- Comfort in front of operators and executives: you can explain a system simply and take feedback well
- Willingness to travel to client sites every few weeks, including internationally
- Bonus: ERP, CRM or supply chain domain experience; Mandarin or German
Engineering Full-Stack Engineer, Operator Workbench San Francisco · Singapore
The workbench is where operators meet the systems we build: not another dashboard, but the surface where a person sees what an agent did, why it did it, and holds the authority to approve, hold or redirect it. You will own that surface end-to-end across a TypeScript and React frontend and the Python services that orchestrate the agents.
What you will do
- Own features end-to-end. Take entire features from design through deployment, across frontend, backend and everything between, and be accountable for how they perform in production.
- Make dense data usable. Build interfaces that turn ERP, CRM and document data into something a buyer, planner or claims handler can act on every day.
- Power the action layer. Design the APIs and services that drive agent orchestration, approvals and write-back, with the reliability others can build on.
- Work with the people using it. Sit with client teams, see where their work breaks down, and ship the fix quickly.
- Raise the bar. Improve developer experience, testing and CI so the whole team ships faster and safer.
What you will bring
- 4+ years building production web applications, owning features from architecture to launch
- Proficient in TypeScript, React and Python; comfortable with PostgreSQL or a similar relational database
- Have mentored engineers or set technical direction on a team
- Care deeply about product quality and how work feels to the person using it
- Bias toward shipping and iterating with users over perfecting in isolation
Engineering Infrastructure and Security Engineer San Francisco · Singapore · Shenzhen
Every client runs in their own tenant, in their own region, under their own data rules. You will build and run the platform our systems live on: deployment pipelines, orchestration, observability, and the security and compliance controls that regulated clients, and the private equity funds that own them, demand.
What you will do
- Own the platform. Design and run infrastructure on AWS, GCP, Azure and regional clouds that supports agent workloads reliably and cost-effectively, one tenant per client.
- Automate deployment. Build the CI/CD, containers and infrastructure-as-code that let a small team stand up a new client environment in days.
- Make everything observable. Logging, monitoring, alerting and an audit trail with sources across distributed agent systems, so we see issues before clients do.
- Meet enterprise standards. Own permissions, data residency, secrets, and the controls behind security reviews in regulated sectors.
- Scale what works. Continuously improve performance, cost and reliability as the number of live systems grows.
What you will bring
- 3+ years in infrastructure, DevOps or platform engineering, including owning production systems
- Proficient with at least one major cloud, Terraform, Docker and Kubernetes
- Have led incident response or set reliability standards for a team
- Understand networking, identity, encryption and compliance fundamentals; SOC 2 or ISO 27001 experience is a plus
- Comfortable in Python and shell; you enjoy building tooling that makes everyone faster
Engineering Data and Integration Engineer San Francisco · Singapore · Shenzhen
Every system we build starts from the same model of the business, drawn from the systems the client already runs. You will connect ERPs, CRMs, PLMs, mailboxes, documents and spreadsheets into that model, keep it correct as the data changes, and make sure gated actions write back cleanly.
What you will do
- Connect the systems. Build and maintain integrations with SAP, Microsoft Dynamics, NetSuite, Salesforce, HubSpot and the long tail of in-house tools, through APIs, files and databases.
- Model the business. Translate customers, orders, parts, invoices, suppliers, contracts and shipments into typed objects, properties and links that every use case can share.
- Make data trustworthy. Build the pipelines, checks and reconciliation that keep the model in step with the source systems, with lineage back to every record.
- Own write-back. Design the gated paths by which approved actions return to the systems of record, safely and auditably.
- Ship with the deployment teams. Work alongside forward-deployed engineers on live client environments.
What you will bring
- 3+ years in data or integration engineering with production pipelines you owned
- Strong SQL and Python; experience with at least one ERP or CRM integration in anger
- Comfortable with messy, partial, multilingual enterprise data and the people who own it
- Rigour about correctness, idempotency and auditability
- Bonus: knowledge graphs or ontology modelling; dbt, Airflow or similar tooling
Research AI Research Engineer, Agents and Evaluation San Francisco · Singapore
You will work between research and production: building the evaluation harnesses, retrieval, fine-tuning and inference systems that move ideas from notebook to a client's live operation, and feeding what happens in the field back into the next round of improvement.
What you will do
- Bridge research and production. Turn promising techniques into robust capabilities that clients depend on: extraction, matching, drafting, negotiation and forecasting agents.
- Make evaluation fast. Build the harnesses that measure whether a change actually improves quality on each client's own history, and hold every model to a weekly floor.
- Build the training stack. Own the data pipelines, fine-tuning and experiment infrastructure that let the team iterate quickly.
- Optimise for production. Drive down latency and cost across model providers without sacrificing reliability.
- Learn from the field. Turn real usage and failures back into research questions.
What you will bring
- 3+ years building production ML or AI systems, owning models from prototype to deployment
- Hands-on experience with LLMs: fine-tuning, retrieval, tool use and evaluation
- Fluent in Python and modern ML tooling (PyTorch, Hugging Face or similar)
- Rigorous about measurement: you measure before you optimise
- Excited by real operating problems, not only benchmarks
Research Applied Scientist, Forecasting and Optimization San Francisco · Singapore
Many of the results we are measured on are numbers: inventory turns, quote win rates, days to cash, negotiated savings. You will build the forecasting, matching and optimisation models behind those numbers, and the scenario tools that show a decision's effect before anything writes back.
What you will do
- Model the operation. Demand and supply forecasting, inventory policy, pricing and quoting, supplier negotiation floors, and matching problems across finance and procurement.
- Ship into the loop. Package models so agents and operators can use them daily, with the confidence and explanation each decision needs.
- Prove the result. Design the baselines, holdouts and measurement that let a client's own numbers decide what scales.
- Build scenario tools. Show what a decision would do across the model of the business before it is taken.
- Publish internally. Turn each client engagement into reusable methods for the next.
What you will bring
- Advanced degree or equivalent experience in operations research, statistics, econometrics or a related field
- 3+ years applying forecasting or optimisation in industry, with results you can point to
- Strong Python; experience with OR-Tools, Gurobi, Pyomo, statsmodels or similar
- Able to explain a model and its limits to a CFO or a plant manager
- Bonus: supply chain, pricing or working-capital domain experience
Apply
Show us your work.
We do not ask for a resume. We would love to see what you have built. Send a link to your portfolio, your code or the systems you have shipped, and a few lines on why this role.