Hello, I'm

Abhigyan

Senior Software Engineer

Building AI/LLM Systems & Scalable Distributed Solutions

Get In Touch

About Me

I architect and build AI-powered platforms that handle millions of requests in production. At Walmart, I lead the AI platform team — building RAG systems, multi-agent orchestration frameworks, and LLM-based developer tooling used by engineers across the org.

My work sits at the intersection of traditional distributed systems and cutting-edge AI infrastructure. I care deeply about building things that scale, are observable, and actually make engineers more productive.

I have 5+ years of experience across fintech, gaming, and enterprise AI — from scaling a platform from 0 to 10K daily users as a founding engineer, to optimizing high-throughput workflow APIs serving 1500+ concurrent executions.

Education

IIT (B.H.U), Varanasi

B.Tech (2016–2020)

Location

India

Experience

5+ Years

Software Engineering

Technical Expertise

Backend & Distributed Systems

Spring Boot Dropwizard Kafka Workflow Engines REST APIs

Cloud & Infrastructure

AWS Docker Kubernetes CI/CD Jenkins

AI & LLM Systems

RAG Multi-Agent Milvus Vision LLMs Claude Code Cursor

Data & Observability

MySQL ELK Stack SonarQube Big Data

Experience

Software Engineer 3

Walmart
May 2024 – Present
  • Reduced P95 latency of core workflow APIs from ~200ms to ~130ms, scaling to 1500+ concurrent executions/sec through DB query optimization and infra tuning
  • Designed multi-tenant onboarding architecture supporting multiple teams and workflow engines per tenant within a shared cluster
  • Built production AI platform: RAG pipelines (Docling, pdfplumber, Vision LLMs), hybrid retrieval (Milvus semantic + BM25 with RRF + cross-encoder re-ranking), LLM response generation with eval pipelines
  • Implemented stateful multi-agent orchestration with task decomposition, parallel execution, dynamic model selection, and deterministic pause/resume for human-in-the-loop code review cycles
  • Built AI-powered developer tooling and introduced spec-driven development workflows using LLMs across the backend team

Software Engineer 2

Rakuten
May 2023 – May 2024
  • Designed high-volume payment batch processing systems and APIs enabling third-party integrations across multiple Rakuten brands
  • Built Java-based mock server framework simulating external payment providers — improving integration test coverage and reliability
  • Owned end-to-end payment service reliability and performance optimization

Founding Engineer

Podeum
July 2022 – May 2023
  • Built and scaled backend services from scratch (Java/Spring) — grew from 0 to ~10,000 daily active users
  • Designed core platform: authentication, virtual economy, in-app purchases, and real-time live score feeds
  • Led technical decisions across the stack as the first engineering hire

Software Developer

Morgan Stanley
Oct 2020 – June 2022
  • Developed and maintained Big Data applications for financial data processing pipelines
  • Set up CI/CD pipelines with Jenkins and Sonar — improving code quality and deployment workflows
  • Automated QA processes reducing manual overhead across the team

Featured Work

Production AI Platform

End-to-end RAG platform at Walmart: Docling + pdfplumber for document parsing, Vision LLMs for OCR, Milvus hybrid retrieval (BM25 + semantic with RRF & cross-encoder re-ranking), and LLM-based response generation with evaluation and observability pipelines.

RAG Milvus Vision LLMs Python

Multi-Agent Orchestration

Stateful AI agent coordination system featuring task decomposition, parallel execution, dynamic model selection, and deterministic pause/resume — enabling human-in-the-loop code generation and review cycles in production.

Multi-Agent Orkes Workflow State

Workflow Engine APIs

Core Start/Update Workflow APIs serving 1500+ synchronous executions/sec with P95 latency reduced from ~200ms to ~130ms. Multi-tenant architecture with shared cluster isolation for multiple teams and workflow engines.

Spring Boot Kafka Multi-tenant

Gaming Platform Backend

Full backend for gaming platform at Podeum: authentication, virtual economy, in-app purchases, and real-time live score feeds — scaled to 10K DAU from day one.

Java Auth Real-time

Payment Processing System

High-volume payment batch processing system at Rakuten with Java mock server framework for simulating external providers — improving integration test reliability across multiple brands.

Java Batch Processing Fintech

AI DevTooling & Workflows

Internal AI-powered developer tooling and spec-driven development workflows using LLMs — improving engineering velocity and consistency across feature development at Walmart.

LLMs CI/CD Developer Experience

Let's Connect

I'm always open to discussing new opportunities, interesting projects in AI infrastructure and distributed systems, or just geeking out about LLMs and multi-agent architectures.

abhi.gyan.met16@itbhu.ac.in github.com/abster12 LinkedIn