Home

Enterprise architect,
twenty-odd years of AWS, Oracle, and the parts in between.

I'm Satish. I hold the AWS Golden Jacket (14x currently active AWS certifications) and lead Agentic AI architecture at HCLTech. Most of what I do is figure out how to get an idea from a customer's slide into their AWS account without breaking anything the auditors care about.

ArchitectAI BuilderCertifiedMentor
Based in/India
Doing/Agentic AI at HCLTech

How it started

I got into IT in 2003 writing PL/SQL at Path Infotech. The first thing I built was a resource-management tool nobody asked for. It kept working, which turned out to be a useful pattern for the rest of my career.

A year later I was on ABN AMRO production databases — 24×7, terabytes, a J2EE app I did not fully understand. Then Tricore, remoting into 15 US clients whose names I still recognise on billing invoices. Oracle was what paid. Oracle was what taught me.

I spent most of 2005–2013 at HCL Comnet and HCL Great Britain, moving between Oracle DBA, Apps DBA, and Track Lead roles across Ford in Chennai, Sensata in the US, Dixons in the UK, and News International in London. The News International project moved Oracle from Solaris to SUSE Linux with Data Guard for DR. Almost nothing about that migration went exactly to plan and it still worked.

After that, three big migrations back-to-back — Exadata for DNB Bank in Norway, two Tieto data centres for Länsförsäkringar in Sweden, and 500-plus Oracle / EBS / PeopleSoft databases for Adeccoacross EMEA. I got good at wave plans and cutover runbooks. Mostly because I'd already seen the messy version.

Around 2020 the work moved to AWS. I led APAC and NORAM4 for HCLTech, ran the Rogers Bank → RCI migration on Terraform and serverless, worked with AWS Pro Serv on ExxonMobil using Migration Factory, designed GovCloud for T-Mobile, and built the FinOps posture for Total Energy. Somewhere in there I got the twelfth AWS certification and, with it, the Golden Jacket.

Since early 2024 I've been doing Agentic AI on Bedrock and AgentCore— eight POCs shipped, two production rollouts in flight. The interesting part is not the model. It's the boundary between a probabilistic thing and an audited enterprise account.

I'm also on the IIT Madras Professional Certificate in Agentic AI through mid-2026 — partly because the field moves fast, partly because sitting in a classroom occasionally is good for me.

How I Work

The whiteboard isn't the architecture

The architecture is what survives audit, cost review, and a CAB. Everything else is a nice diagram.

POCs that carry

A demo that can't move into the customer's account without a rewrite isn't a POC, it's a costume. Ship them with the guardrails and cost model on day one.

The boring 5%

Inventory reconciliation, wave plans, cutover runbooks, DR fall-back. That's where migrations go wrong — never the shiny part.

Build people, not just systems

70+ GETs into L2 cloud roles. Architecture that only I can operate isn't architecture.

Off the clock

What I'm actually doing

Coursework for the IIT Madras program, reading the AWS release notes because they change more often than the docs, and answering questions from engineers moving out of ops into cloud.

What I care about at work

The gap between a slide and a running system. Most of the interesting engineering is in there.

Let's connect

Scoping an Agentic AI POC, planning an AWS migration, or want a second opinion on an architecture — happy to talk.

LinkedIn