Vishnu Vardhan

Engineering complex enterprise systems, architecting start-up innovations, and writing code every day.

VishnuVardhan.java

@Engineer(of = "enterprise healthcare")
@SystemsArchitect(building = "start-up initiatives")
@Entrepreneur
public final class VishnuVardhan implements FirstPrinciples {

  @Builds(with = { SpringBoot.class, SpringAI.class,
                  Kafka.class, Postgres.class })
  Product ship(Idea idea) {
    return understand(idea)
        .design(Simplest.thingThatWorks())
        .write(Code.myself())
        .test().deploy();
  }

  @Philosophy
  List<String> onAI() {
    return List.of(
      "AI does not belong to Python.",
      "Spring AI brings it to where the enterprise runs.");
  }
}

Dallas–Fort Worth, Texas

Now

I have spent twelve years building the enterprise platforms that pay healthcare and family benefits to hundreds of thousands of people in the United States. Work at that scale has one rule: a wrong rule or a missed batch window is a public failure. I have led development teams, carried the annual federal policy changes that re-evaluate every active case, and rebuilt a statewide data-governance platform as its sole architect and developer.

Outside that work I design and build systems for start-ups. RaastaPay, an AI advisory platform for Indian road freight, is the current one: Java 21, Spring Boot, Spring AI and Kafka, built alone from the first commit. jamna.nn is my transformer engine in pure Java, written to understand the machinery from the inside.

AI engineering is not a Python monopoly. The systems that carry real money and real records run on Java, and Spring AI puts language models where those systems already live, with typed contracts, transactions, observability and the same test discipline as everything else. That is where I am placing my bet. I am exploring Embabel, Rod Johnson's agentic framework for the JVM built on Spring AI, and preparing to be part of the rise of Spring AI rather than watch it from the sidelines.

I build from first principles and I write all the code. Language models help me move faster; they do not think for me.

Work

2021 – Current

NessSenior Solutions Engineer

Lead developer on a statewide healthcare and benefits platform. I own change requests end to end, from requirements with policy teams to production release. I own the three annual federal policy cycles that re-evaluate every active case. I re-engineered the nightly interface chain to federal and partner systems and restored reliable overnight windows.

2014 – 2021

AccentureApplication Development Lead

Seven years on a statewide healthcare and family-services platform built on the Accenture Public Service Platform, replacing systems the state had run for 24 years. Led a team of developers through delivery, testing and deployment. Rebuilt the state's master person index as sole architect and developer, from program-level risk to meeting requirements, reporting daily to state leadership while the system stayed live.

2014

24x7booking.comSoftware Engineer

Sole developer for the flight-booking module of a live travel platform in Bangalore, from design through production.

2013 – 2014

Schemes CentralSoftware Engineer

Full-stack engineer on a production e-commerce platform: page flows, front-end features, deployments and platform upgrades, all live.

Entrepreneurship

2023 – Current

RaastaPayPrincipal Engineer & Systems Architect

Sole engineer building a founder's AI advisory product for small trucking operators in India. The system of record does the arithmetic; a panel of Claude-backed advisors on Spring AI reads each offer and returns a verdict in the driver's own language. Every AI output is schema-checked, persisted and auditable. Java 21, Spring Boot, Kafka, PostgreSQL, React Native.

Currently building

jamna.nn: A mini Transformer / LLM engine built from scratch in pure Java, no libraries, not even a math one. It learns to sort numbers the way a language model learns to write, by predicting the next token. Ships with an interactive explainer that walks through embeddings, attention and softmax using the real trained weights. MIT.
options-wall-street: Personal tooling for options analysis. Not a product and not financial advice.

Also working in forks of Spring AI and Tribuo, with contributions in progress.