Senior Data & AI Platform professional & Technical Product Owner — Data Platforms, GenAI, Agentic AI, Databricks, AWS, Azure & AI Governance
Turning complex AI & data-platform capability into governed, business-ready outcomes.
I work at the intersection of enterprise data platforms and applied GenAI. By day I drive platform enablement on a GenAI Platform Team in an APRA-regulated environment — Databricks on AWS, Unity Catalog governance, Azure Identity management, foundation-model adoption, and the operating models that make those capabilities safe to consume at scale.
Outside that role I build. This profile collects production-grade reference projects in the agentic AI space — Pydantic AI agents, Model Context Protocol (MCP) servers, FastAPI services, and the observability and evaluation scaffolding around them. The work spans the full arc from business-analysis rigour (BABOK, governance, requirements) through to deployed, type-safe Python.
I rely on Claude Code for the heavy lifting of implementation and stay close to architecture, governance, and product decisions.
| Project | What it is | Stack |
|---|---|---|
| AI Loyalty & Campaign Builder | Agentic loyalty & campaign orchestration layer for mid-market retail. A Pydantic AI orchestrator coordinating five MCP servers (Shopify, Klaviyo, profile, HubSpot, ServiceNow). | Pydantic AI · FastMCP · FastAPI · Supabase · Railway |
| AI Business Command Centre | A multi-channel AI assistant exposing five business tools (pipeline, invoicing, scheduling, status, reporting) to Slack and beyond, with full tracing and an LLM-as-judge eval harness. | Pydantic AI · FastAPI · Upstash Redis · Logfire |
| GenAI Platform Governance | A vendor-neutral reference architecture for access governance on a modern lakehouse — ABAC + RBAC + workspace binding, row/column-level security, audit. | Unity Catalog · ABAC/RBAC · Reference docs |
| Agentic AI Demo Catalogue | A structured catalogue of 35 agentic AI scenarios scored by business value, build effort and uniqueness — a practitioner's library for scoping client work. | Claude Code · MCP · Solution design |
Architecture diagrams, build playbooks and design docs live inside each repository.
Languages & core
Agentic AI & LLM
Data & cloud platforms
Build, deploy & observe
Business analysis & delivery
- Agentic AI engineering — single-agent orchestration, MCP server design (stdio & streamable-http transports), tool calling, conversation memory, model routing and escalation.
- AI governance & risk — access control patterns (ABAC/RBAC), data classification and masking, auditability, and alignment to regulated-industry controls.
- Data platform enablement — Unity Catalog, lakehouse/lakebase patterns, foundation-model APIs, cost attribution and FinOps for AI workloads.
- Product & business analysis — roadmaps, operating models, requirements, BABOK knowledge areas, and stakeholder translation between technical and commercial audiences.
- AI evaluation & observability — LLM-as-judge harnesses, tracing every tool and model call, and treating evals as CI.
- LinkedIn: linkedin.com/in/ashesh-kumar-86a8a74
- Email: asheshcs@gmail.com
- Open to: fractional AI advisory work across insurance, data platforms and SMB.
Built and documented with help from Claude Code. Australian English throughout.