Forward Deployment Engineer

Unisoft Infotech Pte Ltd; Singapore; On-site; SGD 140,000 - 180,000; Full time

Job summary

Unisoft Infotech Pte Ltd is seeking a Forward Deployed Engineer (AI) for the Financial Services sector in Singapore. You’ll be embedded with banks, asset managers or fintechs to build and deploy AI systems inside their environment, starting with a deep dive into a desk’s workflow and delivering an agent- or model-backed solution.

You’ll engage with front-office, ops, risk or compliance teams, deploy production AI across RAG, embeddings, and agent frameworks, and integrate with core banking

Qualifications

• 4+ years shipping production software; strong Python (plus Java or C++ is useful)
• Demonstrable production AI experience — not prototypes. At least one LLM or ML system you built and ran in production
• Hands-on with LLM application engineering: RAG architecture, embeddings and vector stores, tool/function calling, prompt engineering

Responsibilities

• Sit with front-office, ops, risk or compliance teams to find workflows where LLMs move the needle
• Build and deploy production AI: RAG over policy and docs, agent workflows, KYC/credit extraction, model-assisted surveillance
• Integrate against core systems: Temenos T24, Murex, Calypso, FIS, SWIFT/ISO 20022, FIX
• Own evaluation harnesses, guardrails, drift monitoring, HITSD design, latency tuning
• Work inside the bank's controls — change management, model risk governance, audit trails

Skills

Python

Java

LLM deployment

RAG architecture

Data engineering

Tools

LangGraph

LlamaIndex

Kafka

CI/CD

Job description

Forward Deployed Engineer (AI) — Financial Services

Location: Singapore | Perm / Contract | Some regional travel

The role

You’ll be embedded with banks, asset managers or fintechs to build and deploy AI systems inside their environment. Expect to spend your first weeks understanding how a desk, a payments flow or a credit process actually works, then ship an agent, pipeline or model-backed application that changes it. Getting it into production inside a regulated bank is the job — demos are the easy part.

What you’ll do

• Sit with front-office, ops, risk or compliance teams to find the workflows where LLMs and ML genuinely move the needle, and kill the ones where they don't


• Build and deploy production AI: RAG over policy and research documents, agentic workflows for ops and reconciliation, extraction for KYC/onboarding/credit files, model-assisted surveillance and AML triage


• Integrate against core systems — Temenos T24, Murex, Calypso, FIS, SWIFT/ISO 20022, FIX, internal ledgers and mainframe extracts


• Own the unglamorous parts: evaluation harnesses, guardrails, hallucination and drift monitoring, human-in-the-loop design, cost and latency tuning


• Work inside the bank's controls — change management, model risk governance, audit trails, MAS TRM and outsourcing requirements


• Hand over to the client's own engineers so it survives after you leave

What we’re looking for

• 4+ years shipping production software; strong Python (plus Java or Type useful)


• Demonstrable production AI experience — not prototypes. At least one LLM or ML system you built and ran in production, with real users and real failure modes you had to fix


• Hands‑on with LLM application engineering: RAG architecture, embeddings and vector stores, agent frameworks (LangGraph, LlamaIndex or equivalent), tool/function calling, structured output, prompt and context engineering, fine‑tuning where it's warranted


• Practical grasp of evaluation — how you know the system is working, and how you catch it when it stops


• Solid data engineering: SQL, ETL/ELT, Kafka or similar, warehouse/lakehouse patterns


• Cloud (AWS Bedrock / Azure OpenAI / GCP Vertex), containers, CI/CD; comfortable in locked‑down or air‑gapped environments where you can't call a public API


• Working knowledge of at least one financial domain: payments, capital markets, lending, treasury, or regulatory reporting


• Can hold a technical conversation with a quant, a model risk reviewer and a CIO in the same afternoon