Skip to content

Code that ships. Engineering that scales.

Aditya Gupta

👋

I build production-ready, AI-powered platforms — from React frontends to FastAPI backends, deployed on Docker, Kubernetes, and AWS.

Production systems

AI + Backend

Cloud certified

AWS CCP

Available globally

Remote-friendly

What you get

Clean architecture

Modular APIs, typed frontends, maintainable codebases

Deployed & scaling

Docker, Kubernetes, AWS — built for production from day one

Fast iterations

Clear communication, rapid prototypes, no surprises

Career

Experience

Software Engineering Intern

CollegeDilao

Aug 2025 – Jan 2026 Remote
  • Built robust FastAPI and PostgreSQL REST APIs to centralize analytics across three Learning Management Systems (LMS) serving 500+ active students.
  • Optimized a full-stack application built with React and Docker, increasing GitHub engagement tracking efficiency by 30%.
  • Automated custom dashboards and metrics analytics workflows, reducing educator reporting effort by 40%.
FastAPIPostgreSQLReactDockerREST APIsAnalytics
Work

Featured Projects

Skillyn preview

Skillyn

AI-Powered Career Operating System. Analyzes resumes against job descriptions using semantic matching and generates personalized improvement recommendations.

  • +Semantic resume-to-JD matching with cosine similarity scoring
  • +FastAPI backend with async PostgreSQL ORM integration
  • +Docker containerization for consistent development and deployment
  • +React dashboard with real-time analysis feedback
ReactFastAPIPostgreSQLDocker
ChatPro preview

ChatPro

Real-Time AI Chat Platform. Built with React and FastAPI featuring interruption-safe streaming, conversation persistence, and responsive chat experiences.

  • +Interruption-safe streaming with real-time token rendering
  • +Conversation persistence with SQLite and FastAPI async routes
  • +Responsive chat UI built with React and TailwindCSS
  • +Intelligent context management for multi-turn conversations
ReactFastAPISQLite
CloudVault preview

CloudVault

Collaborative AI Workspace. Document-centric workspace platform focused on collaboration, organization, and intelligent information management.

  • +AI-powered semantic file search with vector embeddings
  • +Real-time collaborative editing with WebSocket synchronization
  • +Express.js + MongoDB backend with JWT authentication
  • +Next.js frontend with workspace-level access control
Next.jsNode.jsExpressMongoDB
See More Projects
Problem Solving

Engineering Highlights

Express/Next.js Authentication Redirection Loop Fix

Security & Session Management

Issue Detected

Dashboard interface would lock up in an infinite redirect spinner when a user session expired or a token cookie signature was invalid.

Root Cause Diagnosis

Next.js middleware detected the 'token' cookie existence and routed users to the secure dashboard page, which requested profile info. The backend returned 401 Unauthorized because the token was expired, redirecting the user to '/login'. The middleware intercepted '/login', saw the token cookie was still present in the browser, and redirected back to the dashboard, creating an infinite loop.

Resolution & Impact

Enhanced the Express backend authenticateJWT middleware to immediately invoke res.clearCookie('token') upon token validation or signature mismatch failures. This ensures the invalid cookie is instantly purged from the browser, allowing the client-side router to land cleanly on the login page.

File: src/middleware/auth.tsCode Change
- } catch (err) {
-   res.status(401).json({ error: 'Unauthorized' });
- }
+ } catch (err) {
+   res.clearCookie('token');
+   res.status(401).json({ error: 'Unauthorized. Token cleared.' });
+ }

AI Cosine Similarity Search Scaling Optimization

System Scaling & Optimization

Issue Detected

Evaluating cosine similarity for user queries in-memory blocks the single-threaded Node.js event loop as the total number of files in the database grows.

Root Cause Diagnosis

Iterating over high-dimensional vector embeddings (1,536 dimensions) for all database records in-memory scales at O(N) globally, consuming unnecessary memory and CPU overhead.

Resolution & Impact

Restructured the search pipeline to query and filter active file records by the specific workspace ID at the database level first. This confines the similarity calculations to only the files belonging to the active workspace.

File: src/routes/search.routes.tsCode Change
// Restricting search array at DB layer first
- const allFiles = await File.find({});
+ const workspaceFiles = await File.find({ 
+   workspaceId: req.params.workspaceId,
+   deletedAt: null 
+ });
Tech Stack

The Stack I Ship With

React
Next.js
TypeScript
TailwindCSS
FastAPI
Node.js
Express.js
PostgreSQL
MongoDB
SQLite
AWS
Docker
Kubernetes
GitHub Actions
React
Next.js
TypeScript
TailwindCSS
FastAPI
Node.js
Express.js
PostgreSQL
MongoDB
SQLite
AWS
Docker
Kubernetes
GitHub Actions
GitHub Actions
Kubernetes
Docker
AWS
SQLite
MongoDB
PostgreSQL
Express.js
Node.js
FastAPI
TailwindCSS
TypeScript
Next.js
React
GitHub Actions
Kubernetes
Docker
AWS
SQLite
MongoDB
PostgreSQL
Express.js
Node.js
FastAPI
TailwindCSS
TypeScript
Next.js
React
Credentials

Certifications

AWS Certified Cloud Practitioner
Verified on Credly

AWS Certified Cloud Practitioner

Issued by Amazon Web Services — validates foundational understanding of AWS Cloud, services, security, architecture, and pricing.

  • +Cloud architecture fundamentals and best practices
  • +AWS core services: EC2, S3, Lambda, RDS, IAM
  • +Cost optimization and billing management
  • +Security, compliance, and shared responsibility model
Research

Publications

IEEE ICCSAI 2025

Low-Light Face Detection and Denoising Techniques using Deep Learning Frameworks

IEEE Xplore
IEEE ICCSAI 2025

Low-Light Face Detection and Denoising Techniques using Deep Learning Frameworks

International Conference on Communication, Security and Artificial Intelligence

  • +Deep learning-based face detection under low-light conditions
  • +Comparative analysis of denoising architectures and preprocessing pipelines
  • +Evaluated on benchmark datasets with quantitative performance metrics
  • +Published and indexed by IEEE Xplore Digital Library
Get In Touch

Contact Me

Let's build something amazing together.

Got an idea?
Let's bring it to life.

Whether you have a question, a project proposal, or just want to connect, I'm always open to discussing new opportunities and creative collaborations.

Connect on LinkedIn

adigupta1620
Collaboration

FROM CONCEPT TO CREATION.
LET'S MAKE IT HAPPEN.

Get In Touch