Software Engineer & Open Source Contributor

Saksham
Sirohi

Full-stack engineer with a focus on open-source infrastructure, real-time systems, and LLM tooling. GSoC engineer at FOSSASIA — twice selected, now building spatial WebRTC platforms using Jitsi.

GitHub LinkedIn Get in touch
About

Engineer by craft,
open-source at heart.

I'm a Software Engineer based in Delhi, India, completing my B.Sc. in Programming & Data Science at IIT Madras. My work spans full-stack architecture, real-time platform engineering, and applied ML — with a consistent throughline in open-source contribution.

I've shipped production-grade features across Django backends, Vue 3 frontends, WebSocket services, and CI/CD pipelines — work that's been reviewed, merged, and depended upon by thousands of users globally. I've served as a maintainer on 5+ repositories and spoken at the FOSSASIA Summit about my open-source work.

I approach problems architecturally — thinking past the immediate fix toward systems that stay reliable, readable, and maintainable at scale. Whether it's reducing RAG token usage by 88%, unifying platform-wide rendering, or designing headless media adapters — I aim to leave codebases in better shape than I found them.

Languages & Frameworks

Python TypeScript JavaScript C / Bash Vue 3 Django Flask Pinia

Data & ML

Pandas Scikit-learn Matplotlib LLM / RAG Prompt Eng. SQL / SQLite SQLAlchemy PostgreSQL

Infrastructure & Tooling

Docker CI/CD Redis Celery WebRTC Web Audio API REST APIs Postman

Design & Other

Tailwind CSS Bootstrap Figma wtForms German (Basic) Hindi / English
Experience

Where I've
worked.

FOSSASIA / GSoC
May 2026 — Present
Google Summer Of Code Software Engineer
  • Loungemesh Platform: Designing and building Loungemesh, a proximity-based spatial video lounge for Eventyay using Vue 3 + TypeScript, with a headless Media Service adapter that fully decouples the 2D spatial canvas from lib-jitsi-meet — enabling future media backend swaps with zero upstream impact.
  • Real-Time Proximity Audio: Implementing per-participant audio attenuation via Web Audio API (GainNode/PannerNode) fed by raw WebRTC tracks, bypassing the Jitsi IFrame API entirely for fine-grained volume control driven by real-time coordinate distance calculations.
  • State Architecture: Designing a Pinia-based world state store managing participant coordinates, presence, and reactive UI boundaries; integrating with Eventyay's JWT auth and room configuration APIs, keeping Loungemesh fully stateless on identity under FOSSASIA.
FOSSASIA
Feb 2025 — Present
Open Source Software Developer & Maintainer
  • Schedule Platform Unification: Unified the schedule with server-preloaded payloads, eliminating redundant API calls; added recording attendee tracking, shared rendering across agenda and video, Google Calendar export, and rich search covering tracks, rooms, popularity, and featured speakers.
  • Editor Infrastructure Overhaul: Replaced Quill with Toast UI markdown editor across tickets, talk, and video components; unified static asset delivery, standardised fonts, favicons, and design tokens platform-wide, eliminating a class of cross-component rendering inconsistencies at scale.
  • Maintainer & Code Review: Reviewed 400+ PRs across Django backend, Vue 3 frontend, WebSocket services, and CI/CD pipelines; covering security patches, payment flows, WebSocket race conditions, and accessibility fixes. Mentored new open-source contributors and maintained 5+ repositories.
  • FOSSASIA Summit Speaker: Presented open-source work and platform architecture at the FOSSASIA Summit, one of Asia's largest open-source conferences.
AlignBits
Nov 2025 — Mar 2026
Software Developer Intern
  • LLM Optimisation: Scaled output limits from ~500 to ~2,000 lines and reduced RAG token usage from 40k to under 5k — an 88% reduction — while evaluating 300+ prompts across multiple agentic workflows with maintained accuracy.
  • Groovy Code Intelligence: Built AI-driven pipelines to detect and resolve Groovy runtime and compilation errors, and engineered a workflow that recommends script optimisations to improve overall code quality.
  • Infrastructure & Reliability: Delivered multi-mapper integrations, resilient cURL parsing across edge cases, and structured logging across internal LLM infrastructure for system visibility and debuggability.
FOSSASIA / GSoC
May 2025 — Nov 2025
Google Summer Of Code Software Engineer
  • Backend Engineering: Built end-to-end Django features — one-ticket-per-user quotas, event filtering, hidden-room synchronisation, and organiser creation flows; resolved critical production failures across event loops, 500 errors, room creation lambdas, and timezone handling for orders, filters, and exporters.
  • Full-Stack Architecture Refactor: Replaced fragile client-side Pretalx polling with server-driven schedule delivery, reducing frontend complexity; unified organiser/event routing across Django, Vue Router, and API layers spanning 3+ repositories.
  • Vue 3 Migration: Migrated the schedule editor, CFP workflow, and video webapp from Vue 2 to Vue 3 with TypeScript + Composition API; resolved migration regressions involving routing, schedule balancing, datetime problems, and deleted-room visibility while maintaining full feature parity.
  • DevOps & DX: Improved Docker workflows to preserve SPA assets, resolved WSL/macOS cross-platform issues across multiple repositories, automated CI/CD documentation publishing, and streamlined onboarding for a large open-source platform transitioning toward a plugin-based architecture under FOSSASIA.
Projects

Projects

01
AI HR Helper Application
Led AI module development for an 8-person Agile team. Built a GenAI Resume Parser using strict prompt engineering to extract JSON metadata and calculate a Suitability Score (0–100). Designed an Adaptive Interview System generating context-aware follow-up questions in real time.
LLM Prompt Engineering Python Agile
02
Bank Telemarketing Prediction
Built a Kaggle classification pipeline with imputation, encoding, feature engineering, SelectKBest, and PCA. Stacking ensemble of 6+ models (RF, GBT, SVM, KNN, MLP) with Logistic Regression meta-learner, optimised via cross-validation. Ranked top 15% (200/1,400) on the leaderboard.
Scikit-learn Pandas Ensemble ML Kaggle
03
Orbit Wars — Reinforcement Learning
Kaggle competition entry for the Orbit Wars RL challenge. Built a self-bootstrapping PyTorch agent ("ProducerLite") under 2,000 lines with a full game-state simulation stack — physics-accurate fleet motion, swept-pair collision detection, per-planet garrison projection, and comet path tracking. Ranked top 1,300 of 5,200+ participants.
Reinforcement Learning PyTorch Python Kaggle
04
Bookstore Pricing Optimisation
Used Python for trend analysis, demand forecasting, and pricing optimisation of second-hand books. Presented key revenue insights using ABC analysis and Pareto charts built with Matplotlib.
Python Matplotlib Data Analysis
05
Household Services App
Responsive web app using Flask and Vue.js to facilitate bookings across household services. Utilised SQLite and RESTful APIs for efficient data management, with Redis and Celery for caching and task scheduling.
Flask Vue.js Redis Celery
06
Library Management System
Real-time borrowing and renewal features improving circulation efficiency. Integrated Flask, SQLite, Bootstrap, and RESTful APIs to create a responsive and dynamic library interface.
Flask SQLite Bootstrap REST API
Open Source Activity

Contributions &
community.

Loading contribution graph…
200+
PRs Merged
500+
PRs Reviewed
8+
Open-Source Repositories Contributed
5+
Open-Source Repositories Maintained

Maintainer at FOSSASIA — contributed across Django backends, Vue 3 frontends, WebSocket services, CI/CD pipelines, and documentation. Speaker at FOSSASIA Summit, one of Asia's largest open-source conferences.

GitHub Profile FOSSASIA Org
Education

Academic
background.

Degrees

Indian Institute of Technology Madras
B.S. in Data Science & Applications
2026
Software Engineering Software Testing AI: Search Methods Deep Learning C Programming Market Research Managerial Economics
IIT Madras
Diploma in Data Science
2025
ML Foundations ML Techniques ML Practice Business Analytics Data Science Tools Business Data Mgmt
IIT Madras
Diploma in Programming
2025
DBMS DSA in Python Full-Stack Web Development REST API Design Java & OOP System Commands
IIT Madras
Foundation — Mathematics & Statistics for Data Science
Completed
Linear Algebra Calculus Probability Inferential Statistics Python Programming Computational Thinking

Certifications & Awards

Best Project — Software Engineering
IIT Madras · 2025
FOSSASIA Summit Speaker
FOSSASIA · Open Source Conference
SQL Intermediate Certification
HackerRank
Google UI/UX Design Professional
Google
Orbit Wars — Reinforcement Learning
Kaggle · Top 1,300 of 5,200+ participants
Bank Telemarketing ML — Top 15%
Kaggle / IIT Madras · 200 of 1,400 teams