Forward Deployed Engineer

Job description

About RE:AI RE:AI is Singtel Digital InfraCo's sovereign AI platform — GPU-as-a-Service on NVIDIA infrastructure, a Model-as-a-Service catalog spanning open- and closed-weight models (Mistral, Qwen, DeepSeek, GLM and others), and a Token-as-a-Service layer that gives every application one governed gateway into that entire model catalog — unified API, policy-based routing, scoped credentials, spend controls, and a single audit trail, all running inside Singtel's sovereign Nxera data centres. Customers span government, financial services, healthcare, and enterprise across ASEAN, all of whom need to move from AI pilot to production without re-litigating security and compliance for every model or integration. The role We're building a customer-facing AI engineering function inside Customer Success to turn AI conversations into signed pilots. AI Engineers sit alongside prospective and existing customers — from the first exploratory meeting through to a live pilot — and are the technical reason a customer says yes. The work spans two areas, and you will do both. Applied AI engineering: prototypes and custom demos, reference deployments and proofs-of-concept, GPU sizing and architecture, inference benchmarking and model fit. Business value engineering: TCO modelling and commercial structuring, tender and pre-sales support, CXO and strategic engagements, thought leadership and GTM. You will build, and you will build quickly — but what you build exists to win and shape a deal, not to run in production for three years. Production hardening and steady-state run sit with the AI infrastructure and production team. Roughly half this job is technical judgement and half is communication and commercial reasoning. What you'll do Scope the opportunity. Work with Customer Success and Sales to understand a prospective customer's problem, workflow and constraints — then decide what to build that will move the deal forward. Build the prototype or custom demo that wins the deal. Stand up working prototypes, custom demos and reference deployments on RE:AI — agentic workflows, retrieval, tool use, guardrails — tuned to the customer's actual use case rather than a generic demo. Demonstrate them live wherever possible, not as a slide deck. Size the platform and prove the fit. GPU sizing and architecture, inference benchmarking, and model selection against the customer's latency, throughput and cost targets — and the ability to show your working when their architects ask how you arrived at the number. Support the pilot. Stay technically engaged as a proof-of-concept converts to a pilot: integration guidance, tuning, and being the escalation point the customer trusts — handing over to the AI infrastructure and production team for build and run once the decision is made. Build a reusable asset library. Patterns that show up across three customers should become a reusable demo, template or accelerator, not three bespoke builds. You'll be expected to spot that and push components back into shared pre-sale tooling. Be the technical face of RE:AI. Represent the platform credibly to customer engineers, architects and CXOs — explaining what you built, why, and what tradeoffs you made, in their language. Support tenders and formal pre-sales. Technical content for RFPs, government tenders and compliance responses — solution write-ups, architecture sections and clarification responses, working with bid and commercial teams to deadline. Model the commercial case. TCO modelling, token economics and commercial structuring — quantify what a solution costs to run on RE:AI and what it is worth against a public API or on-premise alternative, in terms a CFO would accept. Contribute to thought leadership and GTM. Reference architectures, customer-facing write-ups, webinars and conference material that make the next deal easier to open — plus internal enablement so Sales can carry the story without you in the room. What we're looking for Genuinely customer-facing. You can run a technical conversation with a customer's engineers directly, hold your position under pushback, and read a room well enough to know when to go deeper and when to simplify. This is the single most important requirement. AI-native. You have personally built with LLMs and agents — tool-calling, retrieval, orchestration, prompting, evaluation — and you keep up with the field because you're interested in it. Certifications alone won't cover this. Excellent communicator, spoken and written. You can present to a mixed technical and executive audience, and write a clear proposal, tender response or follow-up that stands on its own. Your writing will be seen by customers. Commercial and value judgement. You think about business value, not just the software. You can build or reason about a TCO model, explain where cost sits and what drives it, and articulate why a customer should choose this — and you can say when the right answer is a smaller solution. Comfortable with ambiguity and travel. Every customer and every pilot is different. You'll define your own scope more often than you're handed one, and you're fine being on a plane or on-site when the deal needs it. Sound engineering fundamentals. Strong enough in at least one backend stack to build a credible prototype yourself, comfortable with APIs and cloud basics, and able to reason clearly about how a model behaves in production — latency, cost, failure modes. You don't need to have trained a model or run a production platform. Nice to have Product management or product owner experience — the ability to frame a problem in terms of user value and business outcome. Full-stack capability, enough to make a prototype usable end to end without waiting for anyone. Experience responding to enterprise RFPs or public sector tenders, or building TCO and commercial models. Familiarity with GPU infrastructure, inference serving, or benchmarking model performance and cost. Prior forward-deployed, solutions-engineering, or startup generalist engineering experience. Experience in regulated industries (financial services, healthcare, government) — understanding what "production-ready" means when compliance and data residency are non-negotiable. Experience with LLM gateways/routing layers, guardrail/policy systems, or building on top of an internal model-serving platform. A track record of turning one customer's build into a reusable pattern for the next three.