Aditya
Gupta.
Software Engineer & Cybersecurity Specialist engineering fast, resilient, and attack-hardened digital systems.
About
Computer Science engineer and security researcher passionate about building scalable web applications, attack-resilient architectures, and modern AI platforms that solve real-world problems.
I specialize in full-stack engineering with deep focus on defensive application security, cryptographic implementations, and edge computing. From crafting sub-millisecond user interfaces to architecting hardened backend microservices, I build software designed to perform under pressure.
Security Engineer & Full Stack Intern
Software Developer
BS Computer Science
Projects
A showcase of recent engineering projects.
An all-in-one club management ecosystem for fitness and recreational centers, unifying member registration, swimming pool bookings, cafe POS/inventory, attendance tracking, and biometric device sync into a single dark-themed dashboard.
The Challenge
Replace fragmented, disconnected tools with one unified platform that syncs member status in real time across a biometric access system, handles POS inventory deduction, and automates invoicing/messaging — while staying fast with offline-capable, low-latency reads.
The Solution
Firebase Firestore with persistentLocalCache/IndexedDB for sub-millisecond offline-first loads, paired with Next.js Server Actions. A dedicated integration layer syncs member profiles to the AdviceFit biometric cloud. Cloudinary handles assets, and Meta WhatsApp API plus Resend/Gmail SMTP auto-generate communications.
Transforms static study material into interactive, personalized mock tests and visual mind maps using RAG and LLMs — with progress tracking that adapts to your weak points.
The Challenge
Turn unstructured study content into grounded, hallucination-free practice tests and visual concept maps, with a pipeline that can parse documents, understand semantic relationships, and generate custom assessments.
The Solution
Documents are parsed, chunked, and embedded into a Qdrant vector store. A LangChain-orchestrated RAG pipeline queries Qdrant for relevant context and feeds it to Groq's low-latency LLM to generate questions and Mermaid.js mind maps. FastAPI serves the backend on Azure.
A privacy-first PWA that lets any vehicle owner get notified instantly when they're blocking someone's parking — via a printable QR code that never exposes their phone number.
The Challenge
Enable a stranger to alert a vehicle owner to move their car securely and privately without exposing personal phone numbers, while delivering the notification instantly without an app store download.
The Solution
Each vehicle gets a unique QR code. Scanning it fires a push notification via Firebase Cloud Messaging straight to the owner's device. Supabase Realtime keeps the dashboard live, Razorpay handles billing, and the whole thing ships as an offline-first PWA.
A modern dashboard that lets users create projects, generate embeddable feedback widgets, and view incoming user feedback in real time.
The Challenge
Give SaaS builders a way to collect user feedback securely on their own sites with a lightweight embeddable script per project, and a fast, filterable view of incoming feedback data.
The Solution
Each project gets an auto-generated <script> embed tag. Next.js Server Actions handle backend mutations, Drizzle ORM keeps database interactions type-safe, and TanStack Table powers real-time filtering of feedback data.
An intelligent fashion search system that understands natural language styling queries and recommends compatible products with AI-generated visual pairings.
The Challenge
Move e-commerce fashion search beyond rigid category filters toward genuine style intent — understanding context and visually showing how a product pairs with an outfit.
The Solution
NLP parses natural language queries into style intent. Google Gemini Vision API generates visual previews of outfit pairings, while trend signals are pulled from Google Trends/Instagram. Groq powers fast inference on the Node.js backend.
A modern, responsive converter app for handling unit, currency, and other conversions through a clean, fast, real-time interface.
The Challenge
Build a single lightweight tool that handles multiple conversion categories with instant, accurate results and a UI that feels native on both desktop and mobile.
The Solution
A component-driven React app powered by Vite for fast dev/build cycles. Conversions update in real time as the user types, with source/target selection kept simple and mobile-first.
A React app that helps users discover nearby cafes using the Google Maps Places API — showing them on an interactive map with photos, ratings, and reviews.
The Challenge
Give users a fast, visual way to find cafes near their current location with rich detail (photos, ratings, directions) surfaced through an interactive map and polished popups.
The Solution
Browser geolocation detects the user's position, and Google Places Autocomplete powers location search. Cafe markers render on an interactive Google Map, animating a popup with a photo carousel and reviews.
A lightweight Chrome extension that shows a word's definition and example usage in a popup the moment you double-click it while browsing.
The Challenge
Let users look up unfamiliar words without breaking their reading flow or switching tabs — just a fast popup triggered by a natural browsing gesture.
The Solution
A double-click event listener captures the selected word and queries the Dictionary API for its definition and example sentence, rendering both in a clean, minimal popup.
A decentralized crowdfunding platform where users launch campaigns, fund them with ETH, and funds are strictly released on-chain only if goals are met.
The Challenge
Build a trustless crowdfunding system where fund release isn't controlled by a platform operator but by contract logic itself, verifiable on-chain.
The Solution
A Solidity smart contract deployed on Ethereum Sepolia handles campaign creation, donation tracking, and fund-claiming logic. The React frontend connects via Ethers.js and MetaMask.
A deep learning app that predicts whether a bank customer is likely to churn, using an Artificial Neural Network served through an interactive Streamlit interface.
The Challenge
Turn raw customer profile data into a reliable churn probability using a tuned neural network architecture, made accessible through a no-code interface.
The Solution
A feed-forward ANN trained on ~10K customer records using TensorFlow/Keras. The model and scalers are serialized so the Streamlit app can transform new input and provide real-time predictions.
Services
Technical expertise I bring to product teams.
Secure Full Stack Apps
End-to-end web applications, dashboards, and SaaS platforms built with defense-in-depth architecture, rigorous schema validation, and fluid UI.
AI & ML Platforms
Intelligent applications powered by RAG pipelines, LangChain, and high-throughput LLMs, engineered with strict guardrails against prompt injection.
Hardened APIs & Cloud Systems
High-performance REST & edge APIs, rate-limiting layers, cryptographic authentication, and zero-trust Cloudflare Workers microservices.
Transform Your
Digital Experience
Every system has room to grow. Get a clear view of what works, what holds you back, and how to move toward a secure setup that feels faster, lighter, and resilient under load.
> analyze_architecture()
Analyzing current state & edge posture...
Auditing attack surface [||||||||||] 100%
> harden_systems()
- + Enforce Edge CSP & HSTS Preload
- + Deploy Sub-millisecond Pages Export
- + Attach 10-Layer Bot & DDoS Defense
- + Verify RFC 9116 Disclosure Standard
✓ System optimized & hardened successfully.
Security posture: VERIFIED // Latency: sub-ms
Technical Stack
The technologies and tools I use to build scalable, resilient products.