Sr. Full Stack AIML Engineer

Manager Notes:

• This position will be sitting in India (remote) manager prefers to be located in Bangalore if possible
• Reports to: Innovation team lead / DKB Platform Squad lead
• What success looks like:
  • 3 UC Functions deployed as the DKB Query API accessible via endpoint MCP or A2A ( dkbimpact if Phase 2 triggered)
  • Agent evaluation showing measurable accuracy improvement with DKB context
    10 weekly active queries from real users/agents by Week 14
  • Agent tracing and quality dashboards live in MLflow
• Must haves;
  • AI/ML Applications experience LLM-Based systems Databricks Mosaic AI proficient Python MCP.

What youll do:

• Build the DKB Query API as UC Functions: dkbsearch dkblookup dkbcontent (Phase 1); dkbimpact UC Stored Procedure (Phase 2 if graph extension triggered)
• Implement semantic search over the knowledge graph using Mosaic AI Vector Search
• Expose DKB tools via endpoint MCP or A2A so technical IDEs (e.g. Cursor) agents and functional applications can consume them
• Design and implement agent failure/fallback behavior (empty results stale data traversal timeouts)
• Set up agent evaluation using Mosaic AI Agent Evaluation and MLflow 3.0
• Build agent tracing and observability (query latency accuracy metrics usage dashboards)
• Work with domain users to validate DKB-powered design scenarios (future of supply planning migration assessments architecture Q&A)

Must-have skills:

• 3 years building AI/ML applications with at least 1 year on LLM-based systems (RAG agents tool calling)
• Experience with Databricks Mosaic AI (Vector Search Agent Framework or Foundation Model APIs)
• Python fluency -- building production-quality agent tools not just notebooks
• Understanding of semantic search: embeddings chunking strategies retrieval evaluation (precision recall relevance)
• Experience with MCP (Model Context Protocol) or similar tool-calling patterns
• Comfortable evaluating AI system quality (golden datasets A/B comparison human-in-the-loop review)

Nice-to-have:

• Experience with MLflow (especially MLflow 3.0 agent tracing)
• Experience with UC Functions as agent tools
• Familiarity with technical IDEs (e.g. Cursor) or functional applications that consume agent tools
• Prior work on enterprise knowledge retrieval or domain-specific RAG systems

What success looks like:

• 3 UC Functions deployed as the DKB Query API accessible via endpoint MCP or A2A ( dkbimpact if Phase 2 triggered)
• Agent evaluation showing measurable accuracy improvement with DKB context
• 10 weekly active queries from real users/agents by Week 14
• Agent tracing and quality dashboards live in MLflow