LLM Applications Engineer: Build Production AI Workflows

AZTECH TECHNOLOGIES PTE LTD; Singapore; On-site; SGD 120,000 - 180,000; Full time

Job summary

AZTECH TECHNOLOGIES PTE LTD in Singapore is seeking an AI Application Engineer to translate business problems into production-ready LLM applications. You will design the application logic around the model: prompt architecture, context management, tool calling, RAG workflows, structured outputs, evaluation loops, and user-facing behavior.

You will build AI solutions such as knowledge assistants, document analyzers, and workflow copilots, and implement guardrails, evaluation loops, and user-facing

Qualifications

• Strong Python or TypeScript development skills.
• Hands-on experience with LLM APIs and structured outputs.
• Practical understanding of prompt design and retrieval augmentation.

Responsibilities

• Build AI applications such as knowledge assistants and document analyzers.
• Design prompt systems, retrieval flows, tool calls, and memory.
• Integrate LLM APIs and embedding/vision models with guardrails.
• Create reusable AI workflow patterns for summarization and classification.
• Implement evaluation loops and eligibility checks for high-impact actions.
• Develop testing data and scoring methods for answer quality and latency.
• Collaborate with business users to shape AI product flows with acceptance criteria.

Skills

Python

TypeScript

LLM APIs

Prompt design

Retrieval augmentation

Output validation

Debugging model

Ship usable apps

Tools

LangChain

LlamaIndex

Semantic Kernel

Pydantic

Ollama

Job description

AZTECH TECHNOLOGIES PTE LTD in Singapore is seeking an AI Application Engineer to translate business problems into production-ready LLM applications. You will design the application logic around the model: prompt architecture, context management, tool calling, RAG workflows, structured outputs, evaluation loops, and user-facing behavior.

You will build AI solutions such as knowledge assistants, document analyzers, and workflow copilots, and implement guardrails, evaluation loops, and user-facing