Job Description

Our client is a fast growing Property Tech AI company

About the role

Drive the design, build, deployment, and continuous improvement of agentic AI applications for real estate and construction. You'll own end-to-end agent frameworks, architecting tool chains, crafting prompts, integrating data, and running observability and feedback loops to deliver production-grade AI agents.

Key Responsibilities

• Agent Architecture & Orchestration
  • Design and implement agent pipelines using frameworks like LangGraph, Google ADK, CrewAI, etc.
  • Wire up external APIs, custom tool calls and endpoint integrations
• Prompt Engineering & Feedback loops
  • Develop, iterate, and optimise domain-specific prompt templates at speed
  • Embed complex business logic into multi-step prompt workflows
  • Build user-feedback pipelines to retrain, fine-tune, and adjust agent behavior
• Secure Deployments
  • Containerise and deploy LLMs securely on Azure Cloud (AKS /Azure ML). Familiarity with secure LLM application deployments in Kubernetes clusters.
  • Manage CI/CD for model updates and environment provisioning (Terraform)
• Observability & Evals
  • Build monitoring and logging for agent performance, latency, failures and error handling.
  • Implement agent eval frameworks and track metrics over time (RAGAS, DeepEval/ Langfuse, etc.)
• Collaboration
  • Partner with backend & data engineers to deliver end-to-end product
  • Document architectures, runbooks and learnings from iterations

Required Skills & Experience

• 4-7 years in data science (ML engineering/computer vision) or software engineering, with over 12 months of LLM production experience
• Hands-on with agentic AI frameworks; strong preference for LangGraph
• Expert prompt engineering skills in complex, domain-specific contexts
• Proven track record deploying models in secure cloud environments. Working experience in the Azure ecosystem is essential.
• Experience with containerization (Docker, Kubernetes) and serverless compute
• Familiarity with observability & eval platforms (Langfuse/LangSmith)
• Ability to translate nuanced business logic into detailed agent workflows and prompts.
• Strong coding skills in Python, plus experience with REST/gRPC API integrations

Nice to have

• Fine-tuning open source models (DeepSeek/Llama/Mistral) for narrow domain based tasks, ensuring higher overall accuracy & consistency
• Prototyping RL-based improvements or RL-as-a-Service experiments
• Prior production exposure to real estate or construction domains in building and deploying any data science application (not necessarily LLM based)

They are an early-stage startup, so you are expected to wear many hats, working with things out of your comfort zone, but with real and direct impact in production.

Why our client?

• Fast-growing, revenue-generating proptech startup
• Steep learning opportunities in real world enterprise production use-cases
• Remote work with quarterly meet-ups
• Multi-market, multi-cultural client exposure
• No BS, get-stuff-done culture. Best practical solution wins.