Let's build something great together.§ 01 — Index

MohitKumar

Backend engineer at IIT Jodhpur — I build distributed systems, async pipelines, and high-throughput APIs that hold up under real load. Open to backend roles.

Selected work GitHub LinkedIn LCLeetCode240+Resume

Scroll

02By the numbersMeasured outcomes, not vibes

91.19%

ML accuracy

IMU engagement model · XGBoost + ONNX cross-validation

95.9%

cache hit rate

Redis cache · 24h TTL · payment idempotency proxy

duplicate charges across 20 concurrent requests

Redis SET NX + Lua atomic release · zero race-condition doubles

03AboutProfile

Backend engineer who builds systems that are simple, reliable, and fast.

I design and build backend infrastructure — distributed systems, async pipelines, and payment-grade APIs. Most of my work lives at the intersection of reliability, performance, and clean system design.

Currently building distributed webhook infrastructure with Redis-backed retry scheduling and circuit breaker patterns.

Currently

Distributed webhook infrastructure

Redis-backed retry scheduling · exponential backoff · circuit breaker patterns

04SkillsStack & tooling

Languages

Java
Python
C++
Kotlin
JavaScript
TypeScript
SQSQL
Solidity

Frameworks & Cloud

React
FastAPI
Spring Boot
Redis
PostgreSQL
MySQL
MOMongoDB
Firebase
AWS (S3, EC2)
GCP
REREST APIs

Tools & DevOps

Docker
Git
GitHub
CICI/CD
Linux
Postman
GitHub Actions
Prometheus
Grafana
IntelliJ IDEA
Android Studio
VS Code
Ollama

Systems & Research

learning

DIDistributed Systems
SYSystem Design
CACaching Strategies
SCScalability

05ExperienceSelected roles

Jun 2025 — Jul 2025

IIT Jodhpur

Federated Learning Researcher

Federated learning research with FedMeet, reaching 85% accuracy on meeting client sensor data.
Built an IMU-based engagement system with XGBoost and ONNX, achieving 91.19% cross-validation accuracy.
Integrated the model into backend workflows with communication-aware updates and productivity feedback.

Sep 2025 — Jan 2026

IIT Jodhpur

NLP → Solidity Compiler

Designed a three-phase pipeline that turns natural-language intent into secure Solidity contracts.
Built validated JSON extraction, rule-constrained generation, and automated compilation repair.
Integrated Slither, Mythril, and Semgrep with LLM-based patching for vulnerability detection and fixes.

06Selected workProjects

Source01 / 07

Webhook Delivery System

FastAPIPostgreSQLRedisPython

GitHub

Async webhook delivery engine (FastAPI + PostgreSQL + Redis) returning 202 Accepted on ingest. Models lifecycle as PENDING → IN FLIGHT → DELIVERED → DEAD with a watchdog rescuing stale jobs every 30s via updated-at threshold. Exponential backoff retry scheduler (2s → 32s, max 5 attempts) using Redis sorted set as delay queue. Circuit breaker (CLOSED → OPEN → HALF-OPEN) halting delivery after 5 failures with 60s recovery cooldown. Secures payloads with per-subscription HMAC-SHA256 signatures.

Source02 / 07

Payment Idempotency Proxy

FastAPIRedisPostgreSQLDockerPrometheusGrafana

Stateless payment proxy (FastAPI + Redis + PostgreSQL + Prometheus + Grafana) containerized with Docker Compose — one command deploys all 5 services. Distributed locks via Redis SET NX + Lua atomic release preventing race conditions on concurrent requests; validated 20 simultaneous requests producing exactly 1 transaction with zero duplicate charges. PostgreSQL audit trail (amounts in cents) with Pydantic validation and SHA-256 tamper detection. Redis caching (24h TTL, 95.9% hit rate) with Prometheus metrics exported to Grafana.

Source03 / 07

Spring Email Scheduler

JavaSpring BootQuartzMySQL

Fault-tolerant job scheduling system using Spring Boot and Quartz with MySQL-backed persistence. Supports 50+ concurrent timezone-aware tasks with sub-second execution delays, automatic job recovery, and retry mechanisms.

Live04 / 07

Well-Log-Analyzer

ReactFastAPIPythonPostgreSQLAWS S3Groq

Full-stack sensor data platform processing 50,000+ rows across 100+ channels. Reduced frontend rendering latency by 90% via Intelligent Windowed Downsampling and hit 10,000+ rows/sec backend throughput with async pipelines. Ships a Groq-powered GeoBot (Llama 3.3-70B) with streaming responses and persistent chat history.

Source05 / 07

CloudVault

JavaSpring BootReactFirebaseGCP

Full-stack file management platform with Spring Boot and React. Integrated GCP Cloud Storage and Firebase Realtime DB for per-user metadata. Achieved sub-500ms upload latency for 10MB files under 50 concurrent users. Secured with Firebase ID-token auth and stateless Spring Security.

Source06 / 07

Secure Contract Pipeline

SolidityPythonStatic AnalysisLLM

Automated pipeline for Solidity smart contract generation, vulnerability analysis, and LLM-based patching. Combines static analysis tooling with domain-specific prompting to detect and repair contract vulnerabilities end-to-end.

Research07 / 07

FedMeet

PythonFlowerBiLSTMFederated Learning

Federated learning framework for human activity recognition using multi-sensor IMU data (smartwatch + earables). Implemented gated sensor fusion and BiLSTM to handle non-IID distributions. Achieved 87.97% accuracy — outperforming FedProx, FedPer, and ClusterFL. Co-authored ACM publication.

07ContactGet in touch

Open to backend roles, projects, and good teams.

If you think I might be a fit, feel free to reach out.

mohitkumar4922251@gmail.com

LeetCode240+