Vipul Ponugoti
Currently building at VibeLevel.aiVadodara, India

Vipul Ponugoti Software engineer

Built to
keep going.

Distributed systems. Reliable backends.
I build for the moment something goes wrong.

Explore my work Get my résumé

Java & Spring Boot Python & FastAPI

The resilience lab
A five-node distributed systemFive-node cluster. 5 online. Node A leads. 0 writes committed. Use the node controls below to explore failures and recovery.B / FollowerC / FollowerD / FollowerA / LeaderE / Follower
5 nodes connectedQuorum available · 3 required
Good systems expect failure.

Take a node offline, or send your first write.

Try it yourselfTap a node to stop or restore it
0 committed0 rejected

Cluster ready. A leads; 3 of 5 votes required.

Event log 1 recent event
  1. 01Cluster ready. A leads; 3 of 5 votes required.

Interactive model · simplified elections and instant replication.

SDE Intern at VibeLevel.ai

894 merged pull requests as of 2026-10-21

B.Tech CSE (AI & ML) ’27

Selected engineering

Engineered for
the difficult parts.

Three systems. Three different constraints.
A closer look at the engineering decisions.

Distributed systems Featured project

Raft key-value store

A Raft-replicated key-value store that preserves acknowledged writes through crashes and network partitions.

Java 21gRPCDockerJUnitJMH

20koperations / second

12 msp99 latency

0acknowledged writes lost

Architecture & failure testing
gRPC clientElected leaderMajority replication

The constraint: preserve every acknowledged write when nodes crash or the network splits. The store implements Raft leader election, log replication, and snapshotting behind a linearizable gRPC API.

The test: a five-node cluster and a fault-injection harness covering 50 crash, partition, and message-delay scenarios. Reported leader failover is 350 ms, with no acknowledged writes lost.

The boundary: a five-node Raft cluster requires a majority of three. Losing that majority stops new commits rather than risking inconsistent writes.

Project source is not currently public. Figures come from the project record; hardware, workloads, and raw benchmark reports are not included.

Performance engineering

Exchange matching engine

An exchange matching engine built for sustained throughput, precise execution, and deterministic replay.

Java 21Spring BootKafkaRedisWebSocketJMH

50 µsp99 latency

250korders / second

Architecture & load testing
Order bookKafka event logMarket-data stream

The constraint: match limit and market orders in price-time order while preserving exact deterministic replay through Kafka event sourcing.

Reported measurements include 250,000 sustained orders per second and market-data fan-out to 10,000 concurrent clients, using Redis per-client rate limiting and a custom load generator.

Project source is not currently public. The chart is illustrative; hardware, workloads, and raw benchmark reports are not included.

AI infrastructure

Multi-provider LLM gateway

Multi-provider LLM routing with semantic caching, per-tenant budgets, and failover when a provider goes down.

Spring BootSpring AIKubernetespgvectorRedis

−37%average latency

−36%token cost

Architecture & outage testing
Tenant budgetSemantic cacheProvider routing

The constraint: keep requests flowing through provider failures while controlling per-tenant usage. Spring AI routes requests, Redis enforces rate limits, and pgvector powers semantic caching.

Reported results include 36% lower token cost and a 40% semantic cache hit rate. Simulated outages exercise failover; OpenTelemetry, Prometheus, and Grafana expose system behavior.

Project source is not currently public. Figures come from the project record; hardware, workloads, and raw benchmark reports are not included.

Performance figures are project benchmarks documented in my portfolio, not production guarantees.

Built in the open

Reliability is
a team sport.

My contributions focus on the unglamorous work that makes software safer: authorization, race conditions, input validation, and tests. These are three of those fixes, each linked directly to its merged pull request.

894Merged pull requests
31Repositories

GitHub · 2026-10-21

Explore my contributions
Webhook security

Verify the bytes that were sent.

HMAC verification against the raw request body, preserving the exact payload the sender signed.

Truxify Merged #5296
Concurrency

Make recovery race-safe.

Atomic admission in the circuit breaker’s half-open state, with clamped counter decrements.

ai-code-reviewer Merged #3464
Cache coordination

One miss. One fetch.

Singleflight deduplication lets concurrent cache misses share the same in-flight request.

NexaSphere Merged #3150
Vipul Ponugoti

Based in Vadodara, India

Beyond the code

I like the problems
you can’t see.

A request that races another. A provider that stops responding. A write that must survive a restart. I care about correct behavior and systems you can reason about.

I’m a final-year CSE student and SDE intern at VibeLevel.ai, working across FastAPI and Next.js. Outside that work, I contribute security and reliability fixes to open source.

Currently studying

B.Tech, Computer Science & Engineering (AI & ML)
Parul Institute of Engineering & Technology, Vadodara
2023 – 2027

Always practicing

500+ LeetCode problems
System design & distributed systems
Security & reliability

Experience.

Learning by shipping real work.

VibeLevel.ai

Jul 2026 – present

SDE Intern · Remote

Built an LLM-based human–AI collaboration scoring engine with pytest coverage. Shipped full-stack features across FastAPI and Next.js, hardened MCP contracts, and improved candidate-table performance and authentication reliability.

Explore open-aura

Kshemam Health Solutions

Feb 2026 – Jun 2026

ML Engineer Trainee Intern · Remote

Built Spring Boot APIs and batch pipelines for analytics and model serving. Owned a patient-monitoring risk-classification pipeline, including clinical-threshold tuning, latency benchmarks, and error analysis.

A practical toolkit.

The tools I use to take an idea
from first commit to a working system.

Languages

Java · TypeScript · JavaScript · Python · SQL

Backend

Spring Boot · Spring AI · JPA · FastAPI · Node.js / Express · REST · gRPC · WebSocket · SSE · Microservices · Event-driven design

Frontend

React · Next.js (App Router) · TypeScript · HTML · CSS

Data and messaging

PostgreSQL (pgvector) · MongoDB · Redis · Apache Kafka · Flyway

DevOps and cloud

Docker · Kubernetes · GitHub Actions CI/CD · AWS

Observability and testing

OpenTelemetry · Prometheus · Grafana · Pytest · Vitest · Playwright

Beyond engineering

A practice of
going further.

National awards, shared milestones,
and a curiosity that keeps growing.

National recognition2017 / 2018

National Abacus Championship Gold Topper, twice

34th and 37th National Abacus Competitions, Brainobrainfest 2017 and 2018, Chennai.

National Abacus Gold Topper · 20172017 certificate ↗National Abacus Gold Topper · 20182018 certificate ↗
WISER Summer Program · 2026

3,000 of 65,000 selected

Classical. AI. Quantum.

Exploring classical, AI and quantum approaches to optimization through the WISER Summer Program.

View program certificate

Keep learning. Go deeper.

Selected verifiable credentials

DeepLearning.AI and Stanford Online, via Coursera

Machine Learning Specialization

Verify credential
DeepLearning.AI, via Coursera

Mathematics for Machine Learning and Data Science Specialization

Verify credential
All certifications, programs & coursework Browse the complete collection ↗

Education and credentials

B.Tech, Computer Science & Engineering (AI & ML)

Parul Institute of Engineering & Technology, Vadodara, 2023 – 2027.

Coursework: Data Structures & Algorithms, Object-Oriented Programming, Database Systems, Operating Systems, Computer Networks, System Design, Distributed Systems.

Certifications

Programs

Also completed MIT 6.S191 Introduction to Deep Learning (coursework). Took part in the ISRO Bharatiya Antariksh Hackathon 2026 (certificate), the Cognizant Technoverse Hackathon 2026 (certificate), and the Deccan AI Experts Catalyst Hackathon (certificate).

Start with a conversation

Your next hard problem.
Let’s talk.

A challenging engineering problem, an open-source idea, or a team that cares about the details. I’d love to hear about it.

Start a conversation
vipulponugoti@gmail.com