Forward Deployed Engineer - AI

Sterling Wells Nepal is the operational hub of a UK-based group of chartered certified accountancy firms and fintech companies, including UK Property Accountants, Sterling & Wells, Crane & Partners, FigsFlow, RentalBux, Taxule, and PropertySPV. We support thousands of clients across 50+ countries and oversee accounting for $8B+ in assets.Full-time We are hiring a Forward Deployed Engineer (AI Agents) to sit inside each department of Sterling Wells Nepal, learn how the work is really done, and build AI agents that take on that work safely, with a human reviewer in control. This is a full-time, on-site role in Kathmandu with flexible hours for UK overlap, and you will be the bridge between our accountants, tax teams, marketers, people and growth teams, and product teams and the AI tools they use every day. Your main objective is to sit with each department and produce AI agent that saves real hours on real work. You will build with models such as OpenAI and Claude and connect agents to the systems teams already use, including Xero, Rentalbux, Figsflow, HubSpot, and Microsoft 365. What you’ll own • Map each department's workflows, inputs, outputs, and review points before building anything. • Prioritise opportunities by hours saved, risk, and ease of build, and write a short brief for each agent. • Design and build agents: prompts, skills, playbooks, tool definitions, and MCP servers or integrations. • Build simple internal interfaces so staff can run, check, and give feedback on their agents without engineering help. • Test agents against real (anonymised) past work before release, then track accuracy and reviewer corrections and fix root causes. • Apply data minimisation and least-privilege access to every agent, maintain an agent register, and escalate anything touching regulation. • Pilot with a small group, train staff to use and check agents, and write clear user guides. • Share reusable components across departments, feed insights into the product teams, and report hours saved to leadership monthly. Ongoing success is measured by agents live and in regular use, hours saved per month, accuracy on test sets and reviewer correction rates, staff adoption and satisfaction, and zero data protection incidents caused by agents. What we need • Bachelor’s degree in computer science, Engineering, IT, or a related field, or equivalent practical experience. • 3+ years in software engineering, solutions engineering, or technical consulting, including at least 1 year building with large language models. • Strong Python or TypeScript, with production experience writing clean, tested, version-controlled code. • Experience calling LLM APIs and integrating with REST APIs, JSON, webhooks and OAuth2, including handling rate limits, retries, timeouts and failures. • Hands-on experience with prompt engineering, tool or function calling, agent frameworks, and retrieval-augmented generation (RAG), including vector databases, embeddings and chunking strategies. • Experience building automated evaluation frameworks for LLM agents: test sets, scoring, regression checks, and monitoring or observability of live agents (for example logging, tracing, guardrails). • Comfortable with SQL, Git, Docker and CI/CD pipelines, and at least one cloud platform (Azure preferred). • Able to explain technical ideas in plain English to accountants and non-technical staff. • Fluent written and spoken English is essential. Nice to have • Experience building or connecting Model Context Protocol (MCP) servers, or workflow automation tools such as Make, Zapier, or n8n. • Familiarity with accounting and practice software, for example Xero, QuickBooks, HubSpot, or practice management tools. • Knowledge of data protection rules, including UK GDPR, and experience in an outsourcing or offshore delivery environment serving UK or other overseas clients. • Front-end skills (React or similar) for building simple internal tools. Who you are • Curious about how people work, and patient enough to learn a process before changing it. • Owns outcomes end to end, from first conversation to a working agent in daily use, and pragmatic enough to ship a useful small version first, then improve it. • Careful with client data and honest about what an agent can and cannot do. • Builds trust across teams, handles pushback calmly, writes clearly, and documents as they go. What makes this role different • This is not a build-from-a-distance role; you spend time at the desk with the people doing the work before writing any prompt or code. • You will work across every department of a UK accountancy and software group, from bookkeeping, tax, payroll, onboarding, and AML to marketing, product, support, and HR, with direct access to department heads and UK practice partners. • You will leave each department able to run, monitor, and request changes to its own agents, and feed what you learn into the group's products, including FigsFlow, RentalBux, and PropertySPV. • Every agent you build must protect client data and use only approved AI providers on a least-privilege basis. What we offer • Competitive salary package based on experience and qualification. • Access to current AI models and tools, with a budget for learning and certification, alongside Sterling Wells Academy, LinkedIn Learning, and structured development programmes. • At least one month of total working time dedicated to training and development annually. • Five-day working week, with Saturday and Sunday off. • Daily office lunch, paid annual leave, and sick leave. • Accident and medical insurance coverage. • Regular team-building activities and cultural events. If you want to shape how a growing UK accountancy and software group uses AI, working directly with UK partners, chartered tax advisers, and product leaders, this is your role. Apply with your CV and a short note describing one AI agent or automation you built: the problem, how you built it, and how you measured whether it worked.