Sahil Shinde
Resume (opens in new tab)
SOFTWARE ENGINEER — PUNE, INDIA

Hi, I'm Sahil.I build backend systems and intelligent products.

↳ p95 read latency −18–24% · context precision 0.568 → 0.792

  • Java / Spring Boot
  • React / TypeScript
  • Distributed Systems
  • Applied AI / RAG

I work across Java, Spring Boot, distributed backend systems, React, and measured RAG pipelines. Based in Pune, India, with 2 years of experience and open to globally distributed engineering teams.

Explore my workResume (opens in new tab)
GitHub (opens in new tab)LinkedIn (opens in new tab)
ISelected Work

Selected Work

Backend systems and retrieval engineering, presented through decisions, architecture, and measured outcomes.

I.iBackend Systems / Distributed Systems

Claims Processing System

A Spring Boot claims workflow built around controlled state transitions, secure role boundaries, idempotent operations, caching, and asynchronous status processing.

82.5%

Reporting query latency reduction

58.3 ms10.2 ms

Benchmark result; not production traffic.

  1. Claim request — JWT-authenticated REST request
  2. Spring Security — JWT role-based access control
  3. Claims service — Spring Boot REST API
  4. Lifecycle rules — Controlled state transitions and idempotency
  5. PostgreSQL — Claims data managed with Flyway
  6. Redis — Claim-read cache · linked to Claims service
  7. Apache Kafka — Asynchronous status processing
  8. Status processor — Consumes asynchronous status work
  9. Audit history — Records claim changes
Secure claim requests follow controlled lifecycle transitions, persist to PostgreSQL, use Redis for reads, and publish asynchronous status work through Kafka with an audit trail.
Engineering details
  • JWT role-based access control
  • Controlled claim lifecycle transitions
  • Idempotency and audit history
  • Redis caching
  • Kafka asynchronous status processing
  • Integration testing and CI

~18–24%

p95 claim-read latency improvement

Measured with Redis in a 100K-claim benchmark; not production traffic.

Stack
  • Java
  • Spring Boot
  • PostgreSQL
  • Flyway
  • Spring Security
  • JWT
  • Redis
  • Apache Kafka
  • JUnit 5
  • Mockito
  • Testcontainers
  • Docker
  • GitHub Actions
I.iiApplied AI / RAG

RagLens Evaluation-First RAG System

A retrieval pipeline that narrows, reranks, and evaluates evidence instead of treating a generated answer as proof that retrieval worked.

Engineering details
  • PDF ingestion and chunking
  • 384-dimensional embeddings
  • Qdrant retrieval and metadata filtering
  • MMR candidate selection
  • CrossEncoder reranking
  • RAGAS evaluation
Stack
  • FastAPI
  • LangChain
  • Qdrant
  • HuggingFace
  • CrossEncoder
  • RAGAS
  • React
  1. PDF — Source document
  2. Chunking
  3. Embeddings — 384 dimensions
  4. Qdrant retrieval — Up to 60 candidates
  5. Metadata filtering
  6. MMR — Up to 40 candidates
  7. CrossEncoder — Up to 8 final chunks
  8. LLM context — Final context chunks
  9. RAGAS evaluation — Measures retrieval results
Documents become 384-dimensional embeddings, retrieval produces up to 60 candidates, MMR keeps up to 40, and CrossEncoder reranking selects up to 8 context chunks before evaluation.

0.5680.792

Context Precision

Evaluation result for the RagLens retrieval pipeline.

83.3%100%

Evidence hit rate

Evaluation result for the RagLens retrieval pipeline.

IIProfile

I like understanding what happens between an HTTP request arriving and a reliable response leaving. That leads me into API design, database behaviour, security boundaries, message flows, testing, and performance.

More recently, I've applied the same approach to retrieval and LLM systems, especially the parts that can be measured. React gives me enough frontend range to carry features across the stack.

IIIPrinciples

Engineering is mostly decisions.

  1. III.i

    Design the state transition before writing the endpoint.

    Claims Processing System
  2. III.ii

    Measure before optimizing.

    Claims Processing System
  3. III.iii

    Make failure behaviour explicit.

    Claims Processing System
  4. III.iv

    Authentication is only useful when authorization is correct.

    Claims Processing System
  5. III.v

    Tests should verify boundaries, not implementation details.

    Claims Processing System
  6. III.vi

    An AI answer is not evidence that retrieval works.

    RagLens
IVCapabilities

A practical technical range.

Backend
Java, Spring Boot, REST APIs, Spring Security, JPA/Hibernate
Data & Messaging
PostgreSQL, MongoDB, Redis, Kafka
Frontend
React, TypeScript, JavaScript
Testing & Delivery
JUnit, Mockito, Testcontainers, Docker, GitHub Actions
Applied AI
FastAPI, LangChain, Qdrant, HuggingFace, RAGAS
VExperience

Experience, education, and continued practice.

Avanzens Consultancy Services

Pune

Programmer Analyst (Java Backend Developer)

Sep 2025Present

  • Built Spring Boot services for user management and credit-scoring workflows, with secure REST APIs using PostgreSQL, Flyway, JWT, and Spring Cloud Gateway.
  • Used Kafka for asynchronous credit-score updates, separating core processing from downstream notifications.
  • Measured ~7.4K events/sec median throughput and 39 ms p95 producer-to-consumer latency across 3,000 events in a controlled real-broker benchmark.
  • Tested failure and retry scenarios, and evaluated DLQ handling, consumer idempotency, and database-to-Kafka consistency.

InternMeets

Pune

Software Engineer Intern (Full-Stack)

Mar 2025Aug 2025

  • Built Spring Boot REST APIs and a React + TypeScript UI for patient-record and appointment workflows in a clinic application.
  • Implemented role-based access control for patient, doctor, and admin roles with Spring Security and JWT across API endpoints and frontend routes.
  • Optimized frequently used MySQL queries through query tuning and indexing.

Self-Employed

Pune

Freelance Software Developer

Oct 2023Mar 2024

  • Built ServiceNow workflow automation with Flow Designer, Business Rules, Client Scripts, ACLs, and REST integrations for incident and request management.
VIContact

Have a system worth building? Let's talk.

For engineering roles and thoughtful collaboration, email is the best place to start.

  • Backend
  • Full Stack
  • Applied AI
Sahil Shinde