AI Application Engineer (LLM, RAG & Agentic Systems)

PT Sustainable Living Lab; Kota Bandung; On-site; IDR 400,000,000 - 800,000,000; Full time

Job summary

SL2 Group in Indonesia seeks an AI Application Engineer to build the AI layer of our products: retrieval systems, LLM orchestration, and the evaluation harness that keeps them reliable in production. You will work in a two-person engineering pair with a web developer, start with retrieval and single-turn generation, then extend to multi-step agentic workflows and human-in-the-loop approvals.

Our working language is English; you will collaborate with colleagues in Singapore and with international

Qualifications

• English proficiency required, spoken and written.
• 2+ years building production software with at least 1 year shipping LLM-based apps.
• Strong Python and FastAPI experience.
• Hands-on with LLM APIs (Anthropic/OpenAI/Google).
• Experience with RAG: chunking, embeddings, vector stores, and reranking.
• Experience with tool calling, structured output, and schemas.
• Familiarity with PostgreSQL, Git, Docker, and CI/CD.

Responsibilities

• Design and build retrieval-augmented generation (RAG) pipelines end to end.
• Develop and maintain the LLM orchestration layer and prompts.
• Build and own an evaluation harness for regression testing.
• Design agentic capabilities: planning, memory, and human approval checkpoints.
• Monitor cost, latency, and model routing; enforce ceilings for autonomous loops.
• Expose the AI layer via a documented, versioned API for the web app.
• Document architecture decisions and share knowledge with the team.

Skills

LLM/AI systems

Python

FastAPI

RAG pipelines

Cost & latency budgeting

English proficiency

problem solving

team collaboration

Tools

PostgreSQL

Git

Docker

CI/CD

Job description

AI Application Engineer (LLM, RAG & Agentic Systems)

As an AI Application Engineer, you will build the AI layer of our products: retrieval systems, LLM orchestration, and the evaluation harness that keeps them reliable in production. You will work in a two-person engineering pair with our web developer, who owns the application, while you own the AI service behind it.

The role has a clear trajectory. You will start with retrieval and single-turn generation, where correctness is measurable and the ground is solid. As the product matures, you will extend the same system into multi-step agentic workflows: tool use, planning, background execution, and human-in-the-loop approval. We move in that order deliberately. An agent built on unevaluated retrieval fails in ways nobody can debug.

This is an applied engineering role. You will not train models from scratch. You will make foundation models work correctly, cheaply, and predictably on real data.

Our working language is English. You will collaborate with colleagues in Singapore and with international partners, and you will read documentation, model provider guidance, and research that exists only in English.

Key responsibilities

• Design and build retrieval-augmented generation (RAG) pipelines end to end: document ingestion, chunking, embedding, hybrid search, and reranking
• Develop and maintain the LLM orchestration layer, including prompt and context assembly, structured output, and tool or function calling
• Build and own an evaluation harness: golden datasets, regression testing on every prompt or model change, and failure analysis. Extend it to trajectory-level evaluation as workflows become multi-step
• Design and implement agentic capabilities as the product requires them: tool definitions and schemas, multi-step planning, state and memory across turns, retry and fallback behaviour, and human approval checkpoints for consequential actions
• Instrument and control token cost, latency, and model routing, escalating from cheap to premium models only where quality requires it. Enforce step and spend ceilings on any autonomous loop
• Expose the AI layer as a documented, versioned API consumed by the web application, with clear contracts and error handling
• Implement guardrails against prompt injection, data leakage, and unsafe tool use, and enforce data boundaries for client information. No unattended writes to production systems without an approval gate
• Document architecture decisions and share knowledge across the engineering team

About you

• Professional working proficiency in English, written and spoken
• 2+ years building production software, with at least 1 year shipping LLM-based applications to real users
• Strong Python, including a production web framework such as FastAPI
• Direct experience with LLM provider APIs (Anthropic, OpenAI, or Google) at the API level, not only through high-level wrappers
• Demonstrated RAG experience built from components: chunking strategy, embedding models, a vector store (pgvector, Qdrant, Weaviate, or similar), BM25 or hybrid search, and reranking
• Practical experience with tool and function calling, structured output, and schema enforcement
• Practical experience evaluating LLM systems: building test sets, measuring quality, and preventing regressions
• Working knowledge of PostgreSQL, Git, Docker, and CI/CD
• Comfortable owning cost and latency budgets, and reasoning about tradeoffs between quality, speed, and spend
• Excellent problem-solving skills, with the ability to work independently in a small team

About us

SL2 Group (sustainablelivinglab.org) is an ecosystem of organizations that design and implement solutions to navigate climate, societal, and digital transitions. Founded in Singapore in 2011, we work globally with governments, companies, and grassroots to address issues from climate adaptation and circular economies to green and tech upskilling and community inclusion. By uniting foresight, technology, and community engagement, our collective transforms ideas into scalable impact.