Engineering Leader · Data & AI Platforms

Make the data platform ready for agents.

I build the platform that onboards agentic workloads — conversational analytics and enterprise RAG, running reliably at enterprise scale.

Explore the platform work ↓Get in touch ↗
DATA SYSTEMS • APPLIED AI • PLATFORM LEADERSHIP

About me

Two decades turning
complex systems
into platforms.

I’m Harish Parameswaran, an engineering leader with 20+ years building enterprise cloud and data platforms. I manage global teams focused on bringing agentic workloads onto a unified data and AI platform.

My current work sits where platform engineering meets applied AI: making conversational analytics and enterprise RAG operate with the context, controls, and reliability that enterprise data requires.

A short history of the work ↓

01 / Platform agenda

From model capability
to platform feature.

The work is to make agentic workloads useful inside the governed enterprise data environment teams already depend on.

01

Ask questions in natural language

Conversational Analytics

Onboarding NL2SQL workloads onto the data platform, with execution accuracy improving from 30% to 85%.

Agentic workloadsNL2SQLData platform
↗
02

Retrieve with enterprise context

Enterprise RAG

A versioned semantic extraction engine for enterprise retrieval-augmented generation, built around semantic contracts and an ontology with provenance, confidence, and permission models.

Agentic workloadsSemantic contractsGovernance
↗

How I think about platform progress

A platform
flywheel.

Each workload is a chance to strengthen the shared platform: learn from real use, turn what works into a governed capability, then make the next workload easier to onboard.

01Onboard a workloadStart with a real data and user need
→
02Measure outcomesEvaluate task quality under stated conditions
→
03Encode the capabilityBring context, permissions, and provenance into the platform
→
04Expand adoptionReuse platform capabilities for the next workload

Conversational analytics and enterprise RAG are two current workload paths through this cycle.

02 / Operating context

Build for the
real conditions.

Unified data and AI platform

I manage global teams building a unified data and AI platform that processes 25M+ streaming events a day across 28 regions. The current focus is onboarding agentic workloads: conversational analytics (NL2SQL execution accuracy improved from 30% to 85%) and enterprise RAG through a versioned semantic extraction engine with semantic contracts, ontology, provenance, confidence, and permission models.

Experience20+ years
LocationBay Area, CA

03 / In design

Inference
economics.

A workbench concept for comparing models across practical enterprise workloads, with measured task quality and serving cost as the decision lens.

Read the design overview ↗
WORKLOAD×MODEL×MEASURED OUTCOME
DECISIONS GROUNDED IN RUNS

04 / Career

Experience across
platform eras.

2023 — NOWDirector of EngineeringData and AI platform leadership
2017 — 2023Consulting Engineer / Senior Principal EngineerEnterprise cloud and data platforms
2015 — 2017Senior Manager, Solutions ConsultingMarketo · led 5 solution engineers
2002 — 2015Demo ServicesEnterprise software demonstrations

05 / Areas of focus

Where I can
make a difference.

01Onboarding agentic workloads onto data platforms — the promotion path from provider capability to platform feature

02Conversational analytics (NL2SQL) and enterprise RAG as the two strong agentic motions

03Inference economics — measured accuracy-per-dollar as the way teams should pick models

Enterprise cloud platformsData platformsPlatform onboardingReliabilityConversational analyticsNL2SQLEnterprise RAGSemantic contractsOntologiesProvenanceConfidencePermission modelsResponsible platform adoption

06 / Contact

Let’s build what
comes next.

Open to Director and VP roles in Data & AI Platform leadership.

harishparam@icloud.com ↗

LinkedIn ↗GitHub ↗Bay Area, CA