I build agentic-AI systems that don't just talk about infrastructure — they autonomously plan, reason, and remediate it. Six years turning manual, day-long cloud workflows into safe, audited, sub-ten-minute conversations.
From the support desk to autonomous infrastructure agents.
A mechanical engineer who chose code, then chose to make code that operates the cloud on its own.
I started in production support at a global investment bank, living inside the CLIs and dashboards that keep platforms running. That front-row seat to manual toil is exactly why my work today is automation-first — I'd seen the problem before I built the solution.
Over four years at JPMorgan Chase I moved from supporting applications, to automating them, to engineering the platform itself — leading Terraform upgrades, certificate and secrets automation, and FinOps initiatives across enterprise systems. Most recently I architected TerraDune, an agentic-AI copilot that turns natural language into safe, audited Terraform and AWS actions.
What drives me is the intersection nobody owns well yet: real agentic AI meeting real, regulated infrastructure — agents that ship actual pull requests, not demos.
NV
// Vishnu N V · Bangalore
02 — How I Work
Principles that survive contact with production.
01
Diagnose before fixing
Root cause first — and a list of every affected file before a single line changes. Fewer surprises, cleaner diffs.
02
Confirm the design, then build
Agree on the approach up front. It's far cheaper to move a box on a diagram than to rewrite a merged pull request.
03
Automate the toil
If I've done it by hand twice, I'm building the tool the third time — and teaching it to the whole team.
04
Safety is a feature
Every mutating action is allow-listed, gated behind explicit consent, and fully audited. Autonomy you can actually trust.
An end-to-end agentic-AI platform, delivered as a VS Code chat extension, that turns natural-language intent into safe, audited actions across AWS, EKS/ECS, RDS, Terraform Enterprise, Bitbucket, Spinnaker, Dynatrace and Grafana. A multi-turn Gather→Synthesize loop orchestrates 12+ schema-validated tools; an 8-phase self-healing agent scans modules, rewrites HCL, auto-diagnoses failed plans, re-prompts for fixes, and opens PRs — collapsing 10+ CLIs and 6+ dashboards into one conversation.
A full-stack internal developer portal on Cloud Foundry that catalogs, versions and distributes VS Code extensions and team docs — replacing ad-hoc Bitbucket/email sharing. Its standout engineering is a self-healing persistence pipeline for a stateless PaaS (in-process DB → debounced JSON snapshot → automated Bitbucket commit → Cloud Foundry restage) that guarantees zero data loss across container redeploys.
days→min
Tool publish time
0
Data loss on redeploy
React 18Node.jsExpressSQLiteDockernginxCloud FoundryBitbucket API
03
Pazhamozhi Pattu AI Personal
Personal · Open source
A CLIP-based semantic and visual search engine over a curated catalog — search by meaning or by image, not just keywords. Built with a FastAPI backend and a FAISS vector index, it's my hands-on playground for the same retrieval and embedding techniques I bring into production: a small, real product that doubles as a live demo.
JPMorgan Chase & Co. — Sales & Research Technology, Bangalore
Built TerraDune; led Terraform module upgrades & workspace decommissioning; automated CKMS certificates and SCX secrets rotation; drove S3 FinOps savings and a VPC-cleanup tool recognized by leadership.
2022 — 2024
Production Support Engineer · Associate
JPMorgan Chase & Co. — Bangalore
L1/L2 support across Sales & Research Technology apps in APAC hours; automated installs and ops via Python/ENAML; deployments through Jules & Cloud Foundry.
2020 — 2022
Operations / Support Engineer
Tata Consultancy Services — Client: Deutsche Bank (BFSI)
RPA, risk-analytics and OTP platforms; Ansible & PowerShell automation; Geneos / Splunk monitoring; 100% SLA on production resolution.
2019
Software Engineering Trainee (Python)
ART Technology and Software India Pvt. Ltd.
The pivot — deliberate retraining from mechanical engineering into software.