Deskripsi pekerjaan Junior AI Engineer Pt Cipta Piranti Sejahtera
ACCURATE INDONESIA is looking for a Junior AI Engineer to help us ship Generative AI features that real users depend on. This is a hands-on implementation role, not a research role. You will spend most of your time building AI agents, wiring them into automated workflows, and turning prototypes into services that run reliably in production.
Qualifications
- 0–2 years of professional experience in software engineering, data engineering, automation, or AI/ML implementation. Strong fresh graduates with substantial hands-on Gen-AI projects are welcome to apply.
- Bachelor's degree in Computer Science, Informatics, Information Systems, Engineering, Mathematics, or a related field — or equivalent demonstrable practical experience.
- A portfolio of Gen-AI work you personally built — an agent, a RAG chatbot, an automation workflow, or an LLM-powered internal tool. GitHub repositories, deployed demos, or a walkthrough during interview all count. This matters more to us than years of experience.
Technical Skills - Must Have
- Python — solid working proficiency. You can build and structure a real application, not just run notebooks.
- LangChain — practical experience building chains, agents, tools, retrievers, and memory. You understand what the abstractions do underneath, not only how to copy examples.
- n8n (or an equivalent workflow automation platform such as Make, Zapier, Dify, or Flowise) — you have built multi-step workflows with branching, error handling, and external integrations. Direct n8n experience is a strong plus; if you come from another tool, we expect you to pick up n8n quickly.
- LLM API integration — working with OpenAI, Anthropic, Google Gemini, or similar; understanding of tokens, context windows, temperature, function/tool calling, and structured output.
- Prompt engineering — system prompts, few-shot examples, output schemas, and iterating on prompts based on observed failures.
- RAG fundamentals — embeddings, chunking strategies, vector databases (Pinecone, Qdrant, Weaviate, Chroma, or pgvector), and similarity search.
- REST APIs and JSON — you can read API documentation, authenticate, handle pagination, and debug a failing request.
- Git — branching, pull requests, and collaborative workflow.
- SQL and basic database work — querying, joins, and reading a schema you did not design.
Ways of Working
- Strong debugging instinct. When an agent gives a wrong answer, you can trace whether the problem is the prompt, the retrieval, the tool, or the data — and you do not stop at 'the model is bad'.
- Comfortable with ambiguity. Gen-AI requirements are rarely fully specified; you can propose a concrete approach and validate it fast.
- Ships and iterates. You would rather deliver a working v1 this week and improve it than perfect a design for a month.
- Reads documentation and release notes. This field changes monthly and we expect you to keep up without being told.
- Clear written communication in Bahasa Indonesia and functional English — enough to read technical documentation, follow English-language sources, and write clear documentation.
- Aware of AI risk: data privacy, prompt injection, PII handling, and the limits of what an LLM output should be trusted to decide.
Nice to Have
- Experience with LangGraph or other agent orchestration frameworks for stateful, multi-agent, or human-in-the-loop flows.
- Self-hosting and operating n8n (Docker, queue mode, environment variables, credential management).
- LLM observability and evaluation tooling — LangSmith, Langfuse, Ragas, or DeepEval.
- Backend API development with FastAPI, Flask, or Node.js/Express.
- Docker and basic CI/CD; deploying services to cloud (AWS, GCP, or Azure).
- Running open-weight models locally or self-hosted (Ollama, vLLM) and understanding the cost/quality tradeoff versus hosted APIs.
- Model Context Protocol (MCP), voice agents, OCR / document understanding, or multimodal use cases.
- Fine-tuning, LoRA, or embedding model adaptation.
- Frontend basics (React / Next.js / Streamlit) for building internal tools and demos.
