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 out a Forward Deployed Engineering function inside Customer Success to close the gap between "the demo worked" and "this is running in production for a paying customer." AI Engineers sit embedded with prospective and existing customers — from the first proof-of-concept conversation through a live pilot deployment — and are the technical reason a customer says yes.

This is not a research role and not a pure pre-sales/solutions-engineering role. You will personally write code, build and harden agents and applications, and take work that started as a fast, vibe-coded prototype (the customer's or your own) and turn it into something that survives contact with real data, real users, and a security review. You'll do this on repeat, across multiple customers and industries, working directly against RE:AI's Model Gateway, MaaS catalog, and GPU-as-a-Service platform.

What you'll do

• Pre-sale, technical. Partner with Customer Success and Sales to scope what a prospective customer needs, then build the proof-of-concept or working prototype that turns a sales conversation into a signed pilot — live in front of the customer where possible, not a slide deck.
• Productize prototypes. Take a rough, fast-built prototype — vibe-coded, hackathon-quality, built by us or by the customer — and turn it into something with real error handling, observability, access control, and cost discipline: something that can survive a pilot's real traffic and a customer's security review.
• Build agents and applications on RE:AI. Ship working agentic systems and applications against the RE:AI Model Gateway and model catalog — tool use, retrieval, guardrails, human-in-the-loop escalation — tuned to the customer's actual workflow, not a generic demo.
• Run pilots end to end. Own the technical relationship through a pilot deployment: implementation, integration with the customer's systems and data, performance and cost tuning, and being the escalation point when something breaks in front of the customer.
• Feed the platform. Patterns that show up across three customers become a reusable asset, not three bespoke builds — you'll be expected to notice that and push reusable components back into shared pre-sale and delivery tooling.
• Be customer-facing. You will be in the room (or on the call) with customer engineers, architects, and sometimes executives — able to explain what you built, why, and what tradeoffs you made, in their language.

What we're looking for

• Shipping experience, not just prototyping experience. You've taken something built fast and rough — your own or someone else's — and made it production-credible: better error handling, real auth, monitoring, cost and performance tuning. You know the difference between a demo and a system someone can depend on, and you can close that gap quickly.
• Hands-on agent and application builder. Comfortable building agentic systems (tool-calling, retrieval, orchestration, guardrails) and full applications end to end — backend, integration, enough frontend to make something usable, not just a notebook.
• Comfortable with modern AI-assisted development. You use LLM coding tools and agents as part of how you build — and just as importantly, you know how to take that output and harden it, not just accept it.
• Genuinely customer-facing. You can run a technical conversation with a customer's engineers directly, explain a design decision under pushback, and read a room well enough to know when to go deeper vs. when to simplify.
• 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.
• Solid engineering fundamentals. Strong in at least one backend language/stack, comfortable with APIs, cloud infra basics, and enough data/ML literacy to work knowledgeably with model behavior, prompting, evaluation, and cost/latency tradeoffs — you don't need to have trained a model, but you need to reason clearly about how one behaves in production.

Nice to have

• 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.