CS @ Chitkara University • HackIndia National Champion • Researcher
I engineer high-scale backends and research the math behind machine learning.

• 🎓 CS Undergrad at Chitkara University, Punjab
• 🥇 HackIndia 2024 National Champion (1 of 2,839) · ICIDEA '25 researcher (under review)
• 🛠 Building analytics platforms, fitness apps, SDE prep tools, and Web3 systems that ship to production
• 🔬 Applied ML researcher focused on energy systems and load forecasting
• 🌐 Open to: new engineering partnerships and full-time opportunities

• Languages: Java, JavaScript, TypeScript, Python, SQL
• Backend: Spring Boot, Express.js, Node.js, REST APIs, Microservices
• Frontend/UI: React.js, Next.js (App Router), Tailwind CSS
• Cloud & DevOps: AWS (EC2, S3), Docker, Linux CLI, GitHub Actions
• Databases: PostgreSQL (incl. Neon serverless), MongoDB, MySQL, Prisma ORM
• ML / Data: scikit-learn, Ridge regression, feature engineering
• Auth & Tooling: JWT, Clerk, Git, Postman, Bash

Distributed analytics platform that turns raw Codeforces history into real-time, high-performance insights.

• 🎯 Motivation: Competitive programmers' Codeforces data is locked behind a rate-limited, Cloudflare-fronted API (~5 req/s/IP), making live dashboards slow and brittle. Existing tools re-fetch on every page load.
• ✅ Solution: Decoupled sync architecture — a worker job queue (PostgreSQL-backed) pre-computes per-handle stats; the dashboard reads pre-aggregated rows for sub-second loads. Supports multi-handle linking, rating & tag-mastery overlays, peer comparison, and contest history.
• 📈 Impact: Sub-second dashboard reads even with full contest history; designed to scale by distributing workers across distinct egress IPs to multiply API throughput.
• 🧰 Stack: Next.js 16, React 19, Node.js 22, Express 5, PostgreSQL (Neon), Prisma 6, Docker, AWS EC2, JWT
• 🔗 Repo: https://github.com/Swastik2740S/CODEFORCES_API
• 🌐 Live: https://codeforces-api-rho.vercel.app/auth

Multi-user fitness tracker: TDEE-calibrated macros, 6-day PPL program, progressive-overload logging.

• 🎯 Motivation: Generic fitness apps give every user the same 2,000 kcal plan and ignore equipment, experience, or diet. Lifters either stall on no-progress programs or juggle five separate apps.
• ✅ Solution: 5-step onboarding → TDEE (Mifflin-St Jeor) + activity multiplier → personalised macros and meal plan respecting eggetarian/vegan/high-protein/low-carb preferences → 6-day Push/Pull/Legs program with auto-regeneration → progressive-overload engine that suggests +2.5 kg when rep targets are met. Food search via Edamam with 30-day cache, 7-day backfill for food/steps/weight, timezone-correct "today."
• 📈 Impact: A full plan (macros + 6-day program + 100+ foods) generated in one onboarding flow; macro rings, streaks, and weekly adherence reports keep users accountable.
• 🧰 Stack: Next.js 16, React 19, TypeScript, Clerk, PostgreSQL (Neon), Prisma 7, Edamam API, Zod, Framer Motion
• 🔗 Repo: https://github.com/Swastik2740S/personal_fit
• 🌐 Live: https://personal-fit-nu.vercel.app

Personal, single-user dashboard for DSA, System Design, and Behavioral prep against a one-year plan.

• 🎯 Motivation: SDE candidates juggle Striver A2Z (474 problems), GoF patterns, SOLID, HLD/LLD, and Amazon-style STAR stories across scattered spreadsheets — no single source of truth for "where am I in my prep?" Cloud-based trackers fail offline or during commutes; spreadsheets don't show streaks or pace.
• ✅ Solution: A local-first dashboard tracking DSA (Striver A2Z, 474 problems: E152/M186/H136), System Design (GoF patterns, SOLID, refactoring, HLD, LLD), and Behavioral (Amazon LP STAR stories) against a one-year plan — with a streak heatmap and on-pace indicator. State lives in localStorage under key preptrack-v1; the app is fully offline-capable and zero-setup. Optional email magic-link sign-in enables cross-device sync via Supabase (Postgres + RLS + Realtime), moving only the diff from the seed in indexed deltas — single-digit KB per user.
• 📈 Impact: A 474-problem catalog ships in the bundle, so syncing never uploads the catalog itself. Realtime edits stream across devices in ~1s under RLS isolation. PWA-installable, works offline once visited (service-worker caches the shell), and ships with a backup/restore flow that nags every 7 days. Test seed asserts 474 problems, 191 done, E152/M186/H136.
• 🧰 Stack: Next.js (App Router), TypeScript, Tailwind CSS, Zustand + persist, optional Supabase (Postgres + Auth + Realtime + RLS), PWA service worker
• 🔗 Repo: https://github.com/Swastik2740S/Tracker
• 🌐 Live: https://tracker-nine-lovat.vercel.app

Decentralised, cryptographically verifiable storage for mission-critical space data.

• 🎯 Motivation: Space missions generate petabytes of sensitive data that must remain tamper-proof across multi-decade missions and distributed international teams. Centralised storage is a single point of failure and lacks verifiable integrity.
• ✅ Solution: Files are encrypted client-side (AES-256-GCM via Web Crypto API) before they leave the device → uploaded to Pinata IPFS for redundant global storage → content hash anchored on the ExSat Testnet smart contract with multi-chain verification via Bitcoin OP_RETURN. Includes zero-knowledge proofs (verify existence without exposing content) and is tuned for high-latency deep-space communication. Aligned with NASA CRS-2 data integrity standards.
• 📈 Impact: Tamper-evident storage with cryptographic proof of existence — anyone can verify a record on-chain, but only authorised users can decrypt the underlying file. Conference-selected at Bitcoin 2025.
• 🧰 Stack: Next.js 14, Ethers.js, ExSat Testnet, Pinata IPFS (Filecoin), Web Crypto API, Hardhat, GitHub Actions
• 🔗 Repo: https://github.com/khushsharma509/Immutable-Data-Integrity-Solutions-for-Space-Exploration
• 🌐 Live: https://immutable-data-integrity-solutions-for-space-exploration.vercel.app

Web3-based marketplace for tokenised green credit trading on Ethereum.

• 🎯 Motivation: Carbon-offset / green-credit markets are opaque, with no transparent provenance, high intermediary fees, and weak audit trails. Buyers and sellers can't verify what they're trading.
• ✅ Solution: MetaMask-based onboarding for users → browse tokenised green credits → purchases execute on Ethereum for immutable, on-chain provenance. Built end-to-end in 24 hours during HackIndia 2024.
• 📈 Impact: Won HackIndia 2024 Web3 Track nationally out of 2,839 teams. Demoed end-to-end on-chain credit purchase with transparent transaction history.
• 🧰 Stack: MERN (MongoDB, Express.js, React, Node.js), Solidity, Web3.js, Ethereum, MetaMask
• 🔗 Repo: https://github.com/khushsharma509/HackIndia-Spark-4---Supersqaud
• 🌐 Live: https://anothertry-delta.vercel.app
• 🎥 Demo / Certificate: https://drive.google.com/file/d/1i8qhs4xzMuPRaKeRbYLBJ3ZBYdwktF2p/view
• 📰 LinkedIn Post: https://www.linkedin.com/posts/hackindiaxyz_hackindia2024-hackindia2024-hackathon-activity-7260259500111114240-y2pZ

Scalable, role-based collaboration with cloud-native deployment.

• 🎯 Motivation: Teams need a centralised system for users, projects, teams, and tasks with role-based access control — most off-the-shelf tools either oversimplify permissions or become unwieldy as teams grow.
• ✅ Solution: Spring Boot 3.5 backend with five-tier RBAC (Admin → Project Manager → Team Lead → Member → Viewer), JWT auth with BCrypt hashing, full CRUD across projects/teams/tasks, and a Next.js frontend. Documented via Swagger/OpenAPI, deployed to AWS EC2 via Docker.
• 📈 Impact: Handles 30 requests/second on cloud-deployed infrastructure; production-grade security and audit-ready API surface.
• 🧰 Stack: Spring Boot 3.5, Spring Security, JWT, PostgreSQL, Next.js 13, Tailwind, Docker, AWS EC2, Swagger/OpenAPI
• 🔗 Repo: https://github.com/Swastik2740S/SmartTask

Relational backend for complex organisational hierarchies and timesheet workflows.

• 🎯 Motivation: HR systems with multi-level reporting (employees → skip-level managers) and weekly timesheet approval workflows are typically locked behind enterprise SaaS — there's no clean, open-source backend for educational/startup use.
• ✅ Solution: Express + Prisma + PostgreSQL backend modelling users, roles, departments, positions, employees (with reporting + skip-level managers), and timesheets (daily + weekly) with approval workflow. JWT auth with role-based middleware (Super Admin, Sub Admin, Employee).
• 📈 Impact: Production-shaped HR data model with soft-delete and activation toggles, suitable as a foundation for real HR systems.
• 🧰 Stack: Node.js, Express.js, Prisma ORM, PostgreSQL, JWT
• 🔗 Repo: https://github.com/divanshu-g/Employee-Management-System

Personal portfolio built with Next.js and deployed on Vercel.

• 🌐 Live: https://swastikverma.in

Under review · ICIDEA

• 🎯 Motivation: Standard load models treat weather as a linear covariate, which hides interaction effects — particularly the "Cooling Trap" where rising humidity amplifies cooling load non-linearly.
• ✅ Solution: Modelled peak-demand growth via a Heat × Humidity interaction term (not weather as a linear covariate), capturing the "Cooling Trap" effect. Trained a RidgeCV regression on 12 years of Delhi grid data.
• 📈 Impact: Peak load predicted at R² 0.96 (train) / 0.93 (2024 hold-out) with RMSE 372 MW. Quantified nocturnal load divergence to motivate battery-storage deployment.
• 🧠 Methods: RidgeCV regression, interaction-term feature engineering, time-based hold-out validation

• 🥇 National Champion — HackIndia 2024, Web3 Track (1 of 2,839)
• 🚀 Conference Showcase Selection — Bitcoin 2025 · exSat Track
• 🔬 Researcher (Under Review) — Structural Shift in Peak Power Demand — Delhi Grid (2020–2025), ICIDEA
• 🏅 National Round Finalist — HackFest 2024 · SAP × PSG iTech (Top 3 of 200+)
• 💻 200+ LeetCode problems solved in Java
• 🏅 IEEE Core Member — organized 6+ technical events

Web3 & Blockchain · Distributed Systems · Applied ML · Hackathons

Thanks for visiting! Feel free to connect 💙