Genetic AI Engineer

About the role

You’ll build the AI and machine learning systems that power Matilda Cloud’s intelligent cloud cost optimisation, policy management, and recommendation engine. This means shipping production agentic workflows, RAG pipelines, fine-tuned models, and ML services — all running across AWS, Azure, GCP, and OCI at enterprise scale.

What you’ll work on

• Migrate cloud pricing and policy APIs to agentic AI architectures using LangGraph and MCP
• Build RAG pipelines for AI-generated cost allocation and capacity reports
• Design and execute the distillation pipeline: teacher data capture (Claude/Bedrock) → SFT/DPO training → Qwen student model
• Build the cloud cost forecasting engine and anomaly detection service (Prometheus + time-series ML)
• Instrument LLMOps observability: tracing, cost monitoring, groundedness evals (LangSmith)
• Expose Matilda’s cloud discovery data as MCP tool surfaces for agentic consumption

Must have

• 4+ years of software engineering with Python as primary language
• Hands-on with LLM APIs — AWS Bedrock, OpenAI, or Anthropic — including function calling and structured outputs
• Built at least one production RAG pipeline (chunking, vector DB, hybrid retrieval)
• LangChain or LangGraph experience — agent loops, tool use, state management
• Strong cloud experience on at least two of: AWS, Azure, GCP, OCI — specifically AI/ML services (Bedrock, SageMaker, Azure ML, Vertex AI)
• Kubernetes for deploying and scaling ML services
• Experience with ML data pipelines — feature engineering, ETL for training data
• Time-series analysis or anomaly detection in a production setting

Nice to have

• MCP (Model Context Protocol) — building or consuming MCP servers
• LLM fine-tuning: SFT dataset construction, DPO, or LoRA/QLoRA
• Model distillation — teacher/student patterns
• LangSmith or equivalent LLMOps observability platform
• FinOps / cloud cost domain knowledge (a big plus given our product)
• Experience with LiteLLM, model routers, or multi-provider LLM setups