Data & AI platforms that run themselves.
I architect governed, intelligent data platforms for global CPG and pharma — and bring agentic AI into production, from self-healing pipelines to natural-language analytics.
Databricks & Spark · LangChain · LangGraph · RAG · MLOps · LLMOps · AIOps
Bengaluru, India · 8 years at SigmoidORDER BY share DESC
Years in data & platforms
Engineers led
Interns mentored & assessed
Interns mentored & assessed
Platforms & tools I build with
// About
From hands-on engineering to platform architecture
Eight years ago I started close to the metal — Hadoop clusters, ETL at terabyte scale, 24/7 pipelines. Today I architect the platforms those workloads run on. My work spans the full lifecycle: data engineering, AI/ML enablement, and the operational foundations — DataOps, MLOps, LLMOps, AIOps — that keep enterprise platforms reliable in production.
I focus on architecting enterprise-grade AI solutions and intelligent data platforms on Databricks and Spark for global CPG and pharma clients, while bringing agentic systems into production to drive self-healing, zero-touch operations. My core strength is turning complex data ecosystems into governed, observable, production-ready platforms — and helping the engineers I work with grow alongside them.
// Expertise
AI-first, grounded in real platforms
AI & Agentic Engineering
Production agentic systems — multi-agent workflows on LangChain and LangGraph, permission-aware RAG over regulated knowledge, and AI review agents wired into CI/CD. AI that ships, not AI that demos.
Platform Engineering
Governed, observable lakehouse foundations that pass audit — Unity Catalog, Delta Live Tables, Databricks Asset Bundles, and zero-touch CI/CD across Azure and GCP.
Data Engineering
Pipelines engineered for trust at scale — Delta Lake, PySpark optimization over billions of rows, schema evolution and ACID, and reliable medallion architectures.
Data Operations
Reliability as a discipline — DataOps, MLOps, LLMOps and AIOps: observability, automation, triage, and self-healing operations that cut on-call load and keep platforms dependable.
// How I build
Declarative, governed, repeatable
Platforms defined as code — so environments are reproducible, governed, and safe to ship continuously.
# Databricks Asset Bundle — one definition, every env bundle: name: cpg-intelligent-platform targets: prod: workspace: { host: https://acme.cloud.databricks.com } resources: jobs: self_heal_ingest: { schedule: "@hourly" } pipelines: medallion_dlt: { catalog: unity_prod }
# Multi-agent order resolution — LangGraph from langgraph.graph import StateGraph g = StateGraph(OrderState) g.add_node("intake", intake_agent) g.add_node("validate", validate_agent) g.add_node("resolve", resolution_agent) g.add_conditional_edges("validate", route_exceptions) app = g.compile(checkpointer=traced_store)
-- Governed medallion layer — Delta Live Tables CREATE OR REFRESH STREAMING TABLE silver_sales (CONSTRAINT valid_upc EXPECT (upc IS NOT NULL) ON VIOLATION DROP ROW) AS SELECT * FROM STREAM(bronze_sales) WHERE _ingest_status = 'clean';
// Featured work
Agentic AI, in production
CPG · Agentic AI
Agentic Order Management
A multi-agent system on LangGraph where specialized agents each own a user intent, routed at runtime. It holds context across a conversation, persists session history in a relational database, and runs multi-step workflows through graphs and subgraphs — with every LLM call traced in LangSmith and a CI/CD pipeline automating testing, evaluation, and deployment.
Pharma · Enterprise RAG
Permission-Aware Knowledge Assistant
One conversational interface that lets teams find enterprise knowledge and act on it — retrieve a spec, then raise the Jira ticket referencing it, in a single exchange. Enforces Entra ID SSO, role-based access, and document-level permission filtering on every retrieval, so each person sees only what they are entitled to.
CPG · Self-Serve Analytics
Natural-Language Sales Analytics
A self-serve analytics layer on Databricks Genie that lets brand and category managers ask sales questions in plain English and get governed answers in seconds — built on a semantic layer that gives every team one trusted definition of revenue, volume, and share.
// Contact
Let's build something governed, scalable & smart
Find me
Open to conversations on platform engineering, agentic AI, and data & AI architecture for CPG and pharma.