Skip to content

Backend & Systems Engineer · Pune, Maharashtra, India

I build backends that survive real load.

Backend and systems engineer working in Node.js, TypeScript, Redis and AWS. The problems I like are the ones where correctness and throughput fight each other. Real-time event pipelines. Caching you can trust. Migrations that fail loudly instead of quietly. I run Refactyl, and I am studying B.Tech in Artificial Intelligence.

Yash Sharma, backend and systems engineer based in Pune, India
throughput gain (Express → Fastify)
3.1×throughput gainExpress → Fastify
req/s sustained (benchmarked)
38.9kreq/s sustainedbenchmarked
events/day (event engines)
2M+events/dayevent engines
concurrent users (sub-100 ms)
10k+concurrent userssub-100 ms

What I work on

  • TypeScript
  • JavaScript
  • Python
  • Java
  • Node.js
  • Express
  • Fastify
  • GraphQL
  • WebSockets
  • Redis
  • PostgreSQL
  • MongoDB
  • MySQL
  • Prisma
  • Docker
  • AWS EC2
  • AWS Lambda
  • AWS S3
  • CI/CD
  • Jest
  • Next.js
  • React

Experience

  1. Refactyl

    Founder & Software Engineer

    Dec 2025 — Present

    A developer tool that migrates and refactors production codebases onto modern frameworks. The output has to be a diff someone can actually read.

    • Engineered a deterministic migration engine (Express to Fastify) benchmarked at 3.1× throughput, 12k to 38.9k req/s.
    • Gated every output behind the real compiler (tsc, @vue/compiler-sfc): a file ships only if the compiler accepts it, otherwise it is preserved and flagged rather than silently broken.
    • Node.js
    • TypeScript
    • Docker
    • PostgreSQL
    • Redis
    • AWS
  2. Mighty Champions

    Lead Web Engineer

    Jan 2026 — Present

    Sole engineer for a preventive mental health education nonprofit. The whole web surface is mine, including the tooling behind their fellowship.

    • Shipped 15+ production websites covering six audience programs: college, teens, women, men, 55+, physicians.
    • Engineered the end-to-end admissions platform: marketing site, application flow, enrollment portal, and an authenticated internal dashboard.
    • Next.js
    • React
    • TypeScript
    • Tailwind CSS
    • Vercel
  3. Voxa

    Fractional Tech Lead

    Mar 2026 — Jun 2026

    Engineering lead for an AI powered multilingual speaking coach. I owned the architecture and the team building it.

    • Owned architecture and technical direction while managing a team of 8 engineering interns.
    • Shipped core product features and the AI voice-feedback pipeline; introduced code review, CI/CD and a repeatable release process.
  4. Self-Employed

    Independent Software Engineer

    Mar 2021 — Dec 2025

    Backend work for products that had to hold up under real concurrency, across a long run of client engagements.

    • Architected backend systems for 10k+ concurrent users at sub-100 ms real-time latency, and event engines processing 2M+ events/day with strong consistency guarantees.
    • Cut API response times 40–60% with Redis caching and asynchronous pipelines.

Projects

Technical skills

Languages
  • JavaScript (ES6+)
  • TypeScript
  • Python
  • Java
  • SQL
Backend
  • Node.js
  • Express
  • Fastify
  • REST APIs
  • GraphQL
  • WebSockets
  • Microservices
  • Event-Driven Architecture
Databases & Caching
  • PostgreSQL
  • MongoDB
  • MySQL
  • Redis
  • Prisma
Frontend
  • React.js
  • Next.js
  • Tailwind CSS
Cloud & DevOps
  • AWS (EC2, S3, RDS, Lambda)
  • Docker
  • Vercel
  • CI/CD
  • Jest
  • System Design

Education

B.Tech, Artificial Intelligence

Ajeenkya D Y Patil University, Pune

Aug 2025 — May 2029

Certifications

  • AWSCloud Practitioner Essentials
  • AWSTechnical Essentials
  • AWSGetting Started with DevOps on AWS
  • PostmanAPI Fundamentals — Student Expert

Keep it up

Three servers, and traffic that will not stop rising

Serve requests before they hit the floor. Every hit heats the server that took it, and a server pushed to 100 drops offline. Hammering one lane is how you lose.

It loads nothing until you press Play, so it costs a passing visitor exactly zero.

Requests fall down three lanes. Press 1, 2 or 3 (or tap a lane) to serve one. Every hit heats that server, and a server pushed to 100 drops offline. Spread the load. Three dropped requests ends the run.

Nothing loads until you press start

Building something that has to hold up? Let’s talk.

Email is fastest. I am open to backend and systems work, and to problems that look harder than they should be.