Job description

Design and build full-stack features (web front end, APIs, databases, cloud deployment) for client projects and our own AI/chatbot products.
Build LLM agent harnesses: orchestration and agent loops, tool/function calling, structured outputs, memory and state, and integrations via Model Context Protocol (MCP) servers and APIs.
Engineer prompts and context: system prompts, context assembly and compression, retrieval-augmented generation (RAG) pipelines with chunking, embeddings, hybrid search and re-ranking over vector databases.
Build evaluation harnesses: test datasets, automated and LLM-as-judge evals, regression tests in CI, and quality metrics before and after each release.
Implement guardrails and security: prompt-injection and data-exfiltration defences, tool permissioning and sandboxing, PII masking (PDPA), output validation and human-in-the-loop approval for risky actions.
Run AI features in production: tracing and observability (e.g. Langfuse, LangSmith, OpenTelemetry), model routing, caching, and controlling cost and latency.
Bring AI coding agents (e.g. Claude Code, GitHub Copilot, Cursor) into the team’s development workflow with clear conventions, review practices and guardrails; share knowledge with other developers.
Mentor developers and review code across the stack.

Qualifications

5+ years of professional full-stack development experience.
Strong back-end skills in at least one of Python, Node.js/TypeScript or C#/.NET, and front-end skills in React or Angular.
1+ year building LLM-based features that reached production (not only prototypes), using the APIs of model providers such as OpenAI, Anthropic, Google Gemini or AWS Bedrock.
Hands-on experience with tool/function calling and at least one agent framework or SDK (e.g. LangGraph/LangChain, OpenAI Agents SDK, Claude Agent SDK, Semantic Kernel, LlamaIndex) — and the judgement to know when a simple workflow is better than an agent.
Experience building RAG with a vector store (e.g. pgvector, OpenSearch, Qdrant, Pinecone) and measuring retrieval quality.
Experience writing evaluations for LLM output and using them to make decisions about prompts, models and changes.
Good understanding of LLM security risks (OWASP Top 10 for LLM applications, prompt injection) and how to mitigate them.
Experience deploying to AWS (serverless or containers), with SQL databases (PostgreSQL, SQL Server or MySQL), CI/CD and Git.

Send us your details and resume (PDF) through the application form.