Backend systems.
AI agents.
Building for reality.

I'm Pothuraju Vyshrawan, a final year CSBS student at VIT-AP focused on backend engineering, distributed systems and AI integrated applications.

Explore projects ↗ Download Resume ↓ Let's talk

JavaSpring BootKafkaLangChainLangGraph

ENGINEERING
+ AI

01Distributedsystems

02Agenticworkflows

03Productionthinking

01 / ABOUT

I care about what happens after the demo.

My work sits at the intersection of reliable backend systems and practical AI. I build APIs, event driven services and distributed workflows, then connect them to retrieval, agents and LLM powered features where they genuinely add value.

6Spring Boot services in banking platform

RAGBuilt from first principles

8.01Current B.Tech grade

02 / STACK

Backend

Java · Spring Boot · Spring MVC · REST · Spring Security · JPA/Hibernate · Maven

Distributed

Kafka · Redis · Microservices · SAGA · CQRS · Transactional Outbox · SSE

AI Engineering

LangChain · LangGraph · RAG · Spring AI · prompt engineering · vector retrieval

Cloud & DevOps

Docker · GitHub Actions · AWS EC2 fundamentals · IaC basics · PostgreSQL · ChromaDB

Security

JWT · OWASP Top 10 · input validation · sanitization · secure coding · rate limiting

Testing

JUnit · Mockito · JMeter · Postman · API testing · performance oriented thinking

03 / PROJECTS

DISTRIBUTED SYSTEMSPersonal project

Digital Banking Clearing Platform

Six service Spring Boot architecture with Kafka, Redis and Docker. Uses SAGA orchestration with compensating actions, fraud velocity analysis, OTP step up verification and API gateway rate limiting.

Spring BootKafkaRedisSAGADocker

GitHub repository ↗

SAGA Microservices Architecture Flowchart

Digital Banking Platform SAGA Microservices Architecture Diagram

AI / RAGLive Cloud Deployed

CodeSentinel

Java native agentic RAG system for code understanding. Built the chunking, embedding and semantic retrieval pipeline from scratch with ChromaDB and Groq Llama 3.3 70B.

JavaSpring BootChromaDBRedisDockerRAG

GitHub repository ↗ Build writeup ↗

Java Agentic RAG Pipeline Flowchart

CodeSentinel Agentic RAG Pipeline Architecture Diagram

AUTONOMOUS AGENTS / RAGLangGraph StateGraph

AURA AI

Autonomous research and reasoning platform built with LangGraph StateGraph orchestration. Features hybrid BM25 + ChromaDB dense vector retrieval, an AST inspected Python execution sandbox, and self reflection critic loops with verifiable citation grounding.

LangGraphFastAPIPythonTypeScriptChromaDBDockerHybrid RAGAST Sandbox

GitHub repository ↗

LangGraph Multi Agent StateGraph Flowchart

AURA AI LangGraph Multi-Agent Architecture Diagram

04 / AI LAB

Knowledge on AI: Where I Stand

An honest, transparent map of how much AI knowledge I have: distinguishing between systems I actively build hands on (agents, RAG pipelines, evaluations) versus foundational concepts I understand without pretending to have enterprise work experience.

CORE / ACTIVELY BUILDING

Agents

LangChain tool calling, agent loops, LangGraph state machines, routing, retries, and human in the loop patterns.

HANDS ON / BUILT & TESTED

RAG Systems

Chunking strategies, dense & hybrid vector search, retrieval quality, semantic reranking, context caching, and ChromaDB.

THEORETICAL FOUNDATIONS

Fine Tuning Concepts

Conceptual understanding of PEFT, LoRA and QLoRA, parameter efficiency and quantization trade offs, knowing when fine tuning is needed instead of RAG.

PRACTICAL EVALS

Evaluation & Benchmarks

Evaluating task success, faithfulness, context recall, latency, cost, and regression testing rather than subjective "looks good" checks.

APPLIED PROMPTING

Reasoning & Structures

Few shot calibration, structured JSON outputs, deterministic decomposition, and explicit chain of thought for reliable execution.

PRODUCTION CONCEPTS

Latency & Real Constraints

Understanding production engineering trade offs: token budgets, latency SLAs, streaming responses, timeouts, rate limits, and fallback strategies.

USER

→

ROUTER

→

RETRIEVE / TOOL

→

LLM

→

EVAL

→

RESPONSE

05 / ENGINEERING

Production is a constraint, not an afterthought.

Even though I have not operated enterprise clusters in a full time corporate role yet, I do not build toy demos. I study and build around real engineering fundamentals by designing every service with failure modes, concurrency, latency, and data integrity in mind.

Measure before optimizing

Understanding that profiling real latency, throughput, error rates, and memory usage beats intuition and guesswork.

Design for failure

Knowing distributed networks fail: applying timeouts, exponential backoffs, idempotent consumers, and SAGA compensations.

Security by default

Recognizing that auth, least privilege tokens, input sanitization, secret hygiene, and rate limits belong in the initial design.

Observability & Evaluation

Treating code and AI pipelines with equal rigor: metrics, structured logs, evaluation benchmarks, and deterministic fallbacks.

06 / CONTACT

Have a backend problem
or an AI system to build?

I'm open to software engineering, backend and AI engineering opportunities.

vyshrawanp@gmail.com ↗ Download Resume ↓ LinkedIn ↗ GitHub ↗

Backend systems.AI agents.Building for reality.