Sanjeev Kumar — Application Architect

Designing & Engineering
Distributed Systems
That Scale

Application Architect from INDIA · Java & Spring Boot · Kafka · GCP & AWS
Cloud-native platforms and enterprise AI architecture for banking, insurance and healthcare

View Portfolio
12+
Years Experience
18
Engineers Led
3
Regulated Domains
Scroll to explore
Portfolio

Selected architecture work

The constraints, the tradeoffs I accepted, and what happened next. View full portfolio or open a deep-dive case study below.

Outcome figures on the cards below (for example ~35% incident reduction, ~45% faster design / review cycles) are industry-typical for platforms of this class, labelled the same way in the full case studies — not client-disclosed metrics. Exact volumes remain confidential.

About Sanjeev Kumar

Architecture is a series of tradeoffs

Who: Application Architect. What: distributed Java platforms, event-driven systems and enterprise AI. Expertise and value show up in the tradeoffs I can defend. Full profile · Speaking topics · Free resources

I am Sanjeev Kumar, an Application Architect at Wipro (Technical Architect title) with 12+ years building distributed, cloud-native systems. My work sits where the stakes are highest — transaction platforms that cannot post twice, healthcare journeys that cannot leak, insurance workflows that have to stay auditable for years. I lead architecture across banking, healthcare and retail modernisation programmes, authoring HLD/LLD and ADRs, and setting the design patterns and engineering standards squads build against.

My core is Java and Spring Boot on the back end with event-driven and API-first design — Kafka, Pub/Sub, versioned API gateways — and component-driven React and Angular on the front. I deliver on GCP as my primary cloud, with AWS as a strong second, containerised on Docker and Kubernetes, with CI/CD and observability treated as part of the architecture rather than an afterthought. Notes on architecture topics and enterprise AI architecture live under Architecture.

Most recently I completed an eMasters in AI and Machine Learning at IIT Kanpur, and I apply that directly: retrieval-augmented generation and agentic tooling grounded in real enterprise context, used to accelerate design and delivery without letting a model become the authority. I have led teams of up to 18 engineers, and I still review code. More on background, focus areas and how I work is on the About Sanjeev Kumar page.

End-to-End Architecture

HLD/LLD, ADRs, reference patterns and engineering standards

Event-Driven Systems

Kafka, Pub/Sub, outbox, saga and exactly-once business outcomes

Cloud-Native Delivery

GCP-first, Kubernetes, Terraform, CI/CD and observability

Where I go deep

Backend & Microservices

Java 8–21, Spring Boot, Spring Cloud, JPA, gRPC, REST, API Gateway

Architecture Patterns

DDD, CQRS, Outbox/Saga, Strangler-Fig, Circuit Breaker, Bulkhead, Hexagonal

Cloud & Platform

GCP (primary), AWS, Kubernetes/GKE, Docker, Terraform, Jenkins, Harness, Vault

Security & Compliance

OAuth 2.0, OIDC, JWT, mTLS, RBAC/ABAC, Zero Trust, envelope encryption, KMS

Observability & Reliability

Distributed tracing, structured logging, Prometheus, Grafana, ELK, RTO/RPO, active-active

Applied AI

RAG, prompt engineering, LLM-assisted SDLC, agentic tooling, Copilot (Certified L3)

How I Work

Architecture philosophy

Four things I hold to, whatever the stack turns out to be.

Start from constraints, not from patterns

The interesting question is never "microservices or monolith". It is what the latency budget, the compliance boundary, the team topology and the failure tolerance actually permit. I design from those constraints outward, and I write down the ones I could not satisfy so the next architect knows what I traded away.

Make the right thing the easy thing

Standards that live in a document get ignored. Standards that ship as a reusable library, a service template or a shared UI module get adopted, because they are the fastest route to done. I aim for reference patterns that squads pick up as the path of least resistance rather than as a mandate.

Prefer reversible steps

Large migrations fail at the cut-over, not in the design. I sequence change so every step is small, observable and revertible — façades in front of legacy, incremental extraction, rollback paths rehearsed before they are needed. Slower on paper, faster in practice, because nothing has to be undone under pressure.

Mentoring is part of the architecture

A design only survives if the people building it understand why it is that shape. I run architecture-review forums and code reviews as teaching, not gatekeeping — explaining the tradeoff rather than issuing the verdict. The measure is whether engineers start making the same call without me in the room.

Domain Expertise

Regulated enterprise domains

Where correctness, auditability and uptime are not negotiable.

Banking transaction flow with idempotent posting, ledger entry and audit controls
BFS

Banking & Financial Services

High-volume transaction platforms where a duplicate posting is a correctness failure, not a bug. Strangler-fig modernisation behind API façades, exactly-once posting via idempotency and the outbox pattern, and the security posture a regulated ledger requires — mTLS, RBAC, envelope encryption and immutable WORM audit trails.

Citibank Equifax (DMP) Exactly-once posting Four-eyes / SoD
Insurance policy and claims workflow with rules engine and saga compensation
Insurance

Insurance Platforms

Long-running, rules-driven workflows where state has to stay explainable months after the fact. Java microservices and Kafka-backed processes on AWS, with saga-style compensation for multi-step operations that cannot rely on a single transaction, and security hardening across a multi-team platform.

Xceedance Kafka workflows Saga / compensation 8-engineer team
Healthcare member journeys behind an OAuth-secured gateway with consent and audit services
Healthcare

Healthcare Systems

Member- and provider-facing journeys at national scale, where response time and data handling are both under scrutiny. Refactored hot-path backend workflows to improve p95 API response time by 30–40%, and hardened authN/authZ with OAuth 2.0 and JWT to meet healthcare data-handling expectations.

UnitedHealthcare 30–40% p95 gain OAuth 2.0 / JWT AWS CI/CD
Verified Credentials

Certifications & engagements

Independently verifiable, and named clients rather than logos I do not have permission to use.

Professional Cloud Network Engineer

Google Cloud Certified

Cloud Digital Leader

Google Cloud Certified

Cloud Practitioner (CLF-C01)

AWS Certified · 2023

Solutions Architect — Professional

AWS · SAP-C02

In progress

eMasters — AI & Machine Learning

IIT Kanpur · 2024

GitHub Copilot — Certified (Level-3)

GitHub

Engagement track record

Percentage outcome figures below that also appear on case studies are industry-typical for platforms of this class, not client-disclosed metrics.

Featured on LinkedIn

I write about event-driven design, migration strategy and where applied AI genuinely helps an enterprise SDLC — and where it is still oversold. Recommendations from colleagues and clients are on my profile.

Connect
Recommendations

What colleagues say

Feedback from engineering managers, delivery leads and engineers I have worked with. Individual names and client identities are withheld under engagement confidentiality — contactable references are available on request.

Sanjeev took a payments stack that everyone was afraid to touch and gave us a migration path we could actually execute. He insisted on strangling the monolith service by service behind an API facade instead of a big-bang rewrite, and that decision is the reason we shipped without a single customer-visible outage.

EM

Engineering Manager

Enterprise banking platform, Wipro engagement

Reference available on request

What sets him apart is that he argues about tradeoffs in writing. Every significant call on the healthcare platform came with the option we rejected and why. When auditors asked us to justify the data flow two quarters later, the answer was already documented.

DL

Delivery Lead

Healthcare member platform, Wipro engagement

Reference available on request

We brought him in for the Kafka design and ended up leaning on him for the whole event model. He was pragmatic about it: idempotent consumers and replay first, exotic patterns never. The pipeline has been quiet in production ever since, which is the highest compliment I can pay an integration.

PE

Principal Engineer

Insurance policy and claims systems

Reference available on request

He reviewed my designs without ever writing them for me. I would come with one option, he would send it back asking what breaks at ten times the load. Two years of that changed how I approach system design, and I still use his review checklist with my own team.

SE

Senior Software Engineer

Mentee, cloud platform team

Reference available on request

Hiring and want to speak to someone directly? Ask for references or connect on LinkedIn.

Writing

Notes on architecture

Longer-form thinking on system design and cloud-native delivery.

Microservices design patterns
Architecture

Microservices Design Patterns

Designing microservices architectures that stay scalable, maintainable and resilient as the estate grows.

Read more
Cloud architecture patterns
Cloud

Cloud Architecture Patterns

Cloud-native architectures, serverless patterns and practices across GCP and AWS.

Read more
System architecture best practices
Architecture

System Architecture Best Practices

Architectural patterns and practices for building robust, maintainable systems at scale.

Read more
Contact

Let's talk architecture

Hiring for an architect role, or want a second opinion on a design? I reply within 24 hours.

Email

info@sanjeevk.net

LinkedIn

linkedin.com/in/sanjugen

Location

Delhi-NCR, India
open to Bengaluru, Pune, Chennai & Hyderabad

Phone

Available on request

Message sent

Thanks — I'll get back to you shortly.