Technical Architect · Mindsprint
Solution architecture · Platform engineering · Chennai, India · 16+ years

I build the platforms engineering teams ship on.

Solution architect across React, Python/FastAPI, Azure and Kubernetes. Lately: internal developer platforms, release safety, performance work you can measure, and putting AI to work in the delivery pipeline, with a person still making the call.

Read the Relay case study → View expertise LinkedIn ↗
Solution architecturePlatform & DevOpsAzure · KubernetesPerformanceAI in deliveryTeam leadership
relay / deploys14 Oct – 5 Nov 2026
dev
650
uat
489
prod
98
Hover a bar for detail
0deploys in 3 weeks
0success rate
0services shipped
0engineers deploying
16+ yrsWeb, mobile, TV and platform engineering
97.5% → 99.3%Deploy success rate across the org, one release cycle
12.4 s → 0.64 sSCM 2.0 reload time after a shared Redis cache
300 → 42 KBPlanning Tool route bundle after a performance audit
Expertise

What I bring as an architect

Nine areas I work across, each backed by delivered work. Select Proof to jump to the evidence.

01

Solution & system architecture

Breaking large systems into parts teams can own and ship independently, with reviews that keep the design honest.

Micro frontendsMicroservicesSOLIDMVVMArchitecture review
ProofLed a monolith → micro frontends and microservices migration →
02

Platform engineering & DevOps

Internal developer platforms where the safe path is the default: build once, promote the same image, rules enforced on the server.

GitHub ActionsAzure DevOpsCI/CDRelease engineering
ProofBuilt Relay: 1,237 deploys in 3 weeks at 99.3% →
03

Cloud & Kubernetes

Running production workloads on Azure with right-sized Kubernetes resources and a playbook for when they misbehave.

AzureKubernetesHelmAzure Blob StorageAZ-900
ProofOOM playbook across the Infinity Helm charts →
04

Performance engineering

Measure first, fix second, and prove it with before-and-after numbers from a real environment.

RedisProfilingBundle analysisasyncio
ProofReloads 12.4 s → 0.64 s; bundle 300 → 42 KB; a step 46× faster →
05

Backend & data

Async Python services over large supply-chain datasets, with caching and data stores chosen per workload.

PythonFastAPIasyncioPandasRedisPostgresSnowflake
ProofAudit programme: 23 issues fixed, 1,705 ms → 500 ms →
06

Frontend & mobile

From TV set-top boxes and Android to enterprise React: apps on every screen, architected to scale.

ReactTypeScriptReact NativeAndroidHTML5 / OpenTV
ProofOTT apps at NAGRA; React and mobile architecture at Mindsprint →
07

Security & identity

Authentication and access control built into the platform, not added afterwards.

KeycloakRole-based accessQuality gatesSonar · Checkmarx
ProofCompany-wide Keycloak session library →
08

AI in the delivery pipeline

AI that drafts, reviews and explains, while a person makes the decision.

AI PR reviewGuardrailsMeasured adoption
Proof258 AI-assisted PR reviews in 30 days, 1.6 min median →
09

Technical leadership

Leading teams and clients from requirements to delivery: stakeholders, stories, planning, mentoring.

Stakeholder managementRequirements → storiesAgileMentoring
ProofLed the Infinity SCM team; client on-site planning workshop →

Technology stack

Frontend
ReactTypeScriptES6Micro FrontendsReact NativeAndroid
Backend
PythonFastAPIasyncioPandasRedisPostgresSnowflake
Platform
AzureAzure Blob StorageKubernetesHelmGitHub ActionsAzure DevOps
Practice
Solution designCode & architecture reviewSOLIDMVVMAgileMentoringTeam leadershipStakeholder management
Architecture principles

How I work

Measure, then fix

Every performance change ships with before-and-after numbers from a real environment, not a guess.

Guardrails over gatekeepers

Put the rule in the system so the right thing happens by default, and people spend review time on judgement.

AI advises, people decide

AI drafts, reviews and explains. It never approves or merges. Its usefulness is tracked, not assumed.

Featured case study · Platform architecture · Built end to end

Relay: one console to build, deploy, review and approve.

Relay is the internal delivery console for the Infinity platform at Mindsprint. I designed and built it end to end, from product and UX to backend, GitHub integration, access control and AI review, and run it as a shared service for the teams shipping across 38 repositories.

01 · The problem

Shipping by hand

Thirty-plus services meant juggling GitHub Actions runs, release tags and approvals by hand, with little view of what was actually live in each environment.

02 · The approach

Make the safe path the easy one

Build once and promote the same image. Enforce release rules on the server, not just the screen. Add AI where it saves reviewer time, and keep it advisory.

03 · The outcome

1,237 deploys, 99.3% clean

In three weeks, 32 engineers shipped 33 services through the platform. Production success rose from 95.0% to 96.9%, and AI review grew from 21 to 258 reviews a month.

How a release moves through Relay
overlap lock per service + envserver-enforced tag rulesHelm drift flagged for redeploymissing secret keys checkedevery deploy on record

What I built into it

Two release cycles, side by side

Measure30 Sep – 13 Oct14 Oct – 5 Nov
Deploy success rate, all environments97.5%99.3%
Deploy success rate, production95.0%96.9%
AI reviews completed21 to date258 in 30 days
Engineers deploying2832

Success = succeeded ÷ (succeeded + failed); cancelled runs excluded. Source: GitHub deploy history and Relay's own records.

Selected work

More architecture work

Performance · Redis caching · SCM 2.0

Shared Redis cache: reloads from 12.4 s to 0.64 s

Designed and shipped a shared Redis caching layer for the SCM 2.0 backend: 12 caches with 5 to 60 minute expiry and versioned keys, shared across workers, with simultaneous requests for the same data merged into one query.

Validated against an environment without Redis on two separate dates, with an uncached control endpoint (1.34 s vs 1.42 s) to show the gain comes from caching, not the environment. Released to QA on 11 Nov 2026 with no Redis warnings, no server errors, and all 33 SCM 2.0 pages passing regression.

88%less wait on Funding & Delivery vs no cache (1.23 s vs 10.30 s)
0 / 72slow repeat reads with Redis, vs 37 / 72 without
33 / 33pages clean in regression after release
Redis 7FastAPIRequest coalescingKubernetes
first visit vs reloadQA · 11 Nov 2026
Funding & Delivery12.4 → 0.64 s
BVO Run Check12.7 → 0.40 s
BVO headline figures12.6 → 1.0–1.25 s
first visitreload (from Redis)
Leadership · Project Infinity

Leading the Infinity supply chain team

Led the entire supply chain management team for Project Infinity for about a year: managing stakeholders and turning client requirements into delivery-ready stories. Invited on-site by the client to a workshop to plan the next three months.

Team leadershipStakeholder managementRequirements → storiesRoadmap planning
Performance · SCM platform

Audit-driven performance programme

Audited five supply-chain codebases and drove fixes through with measured before-and-after numbers: 23 issues fixed, Output Summary from 1,705 ms to 500 ms, and a vectorised weather-feature step measured 46× faster with identical output.

FastAPIasyncioRedisReactSnowflake
Frontend · Infinity SCM

Planning Tool, 300 KB → 42 KB

Cut the Planning Tool route bundle by about 86% and moved blocking workbook and proxy work off the event loop, so large run bundles no longer stall workers.

ReactTypeScriptBundle analysis
Reliability · Kubernetes

OOM playbook for Helm workloads

Defined CPU and memory requests and limits across the Infinity Helm charts, with a playbook for diagnosing and preventing out-of-memory kills in production.

KubernetesHelmAzure
Architecture · Mindsprint

Monolith to micro frontends and microservices

Led the end-to-end migration of a large monolithic frontend and backend, giving teams independent deployments and faster releases.

Micro FrontendsMicroservicesCI/CD
Platform · Company-wide

Keycloak authentication library

Built and shipped a custom Keycloak library for user session management that was adopted across the company.

KeycloakReactSecurity
Experience

From TV set-top boxes to cloud platforms.

Sixteen years across supply chain and OTT: native Android and React Native on phones and TVs, then React and Python at enterprise scale, then the platforms underneath.

2024 – now

Technical Architect · Mindsprint+

React/TypeScript with FastAPI and Snowflake; full DevOps lifecycle on Azure DevOps and GitHub Actions; Helm and Kubernetes; built Relay; shipped a shared Redis cache that cut SCM 2.0 reloads from about 12 s to under 1.3 s. Led the Project Infinity supply chain management team for about a year, owning stakeholder management and requirements-to-story conversion, and invited on-site by the client for a three-month planning workshop.
2021 – 2024

Senior Lead Engineer · Mindsprint+

Led React web delivery and all mobile architecture in React Native and Android; company-wide Keycloak library; mentoring.
2014 – 2021

Software Expert · NAGRA+

OTT apps for phones and TVs in React Native and native Android, plus HTML5 apps on the OpenTV OS.
2013 – 2014

Technology Analyst · Infosys+

Hybrid mobile and web apps with Angular and Backbone.js.
2010 – 2013

Project Engineer · Wipro+

Native Android apps integrating the ATSC stack for US customers.

Credentials & recognition

Microsoft Certified: Azure Fundamentals (AZ-900)September 2024
Bravo Award · NAGRAResolving complex customer issues on time for Claro Android TV
Best Team Award · MindsprintDelivering a React/Python project on a tight deadline to coding standards
B.Tech, Computer Science · SASTRA University2005 – 2009 · GPA 8.0 / 10
Contact

Building a platform, or fixing one?

I'm glad to talk about architecture, delivery platforms, performance, and AI in the engineering workflow.

Connect on LinkedIn ↗