Kaushal Vaid
Full-stack software engineer fluent in Databases, APIs and UI. Currently leveling up my skills in AI and Deep Learning to understand how to teach machines to do the heavy lifting.
Finetuning a model until my laptop sounds like a Boeing 747 preparing for takeoff.
About Me
As a Data Science student at IIT Madras and an experienced freelance web developer, I thrive at the intersection of software engineering and artificial intelligence.
My background in building scalable, full-stack platforms provides a strong engineering foundation for my work in machine learning and deep learning. I love integrating Machine Learning and Deep Learning models to create intelligent, data-driven applications with seamless user experience.
I teach as much as I build — TAing Statistics and Core Java Programming at IIT Madras.
Education
BS in Data Science and Applications
Indian Institute of Technology, Madras
XII – ISC
Narbheram Hansraj English School, Jamshedpur
Experience
Open-Source Contributor - Girlscript Summer of Code (GSSoC'26)
GirlScript Summer of Code
Profile- Implemented modular backend features for the career-pilot repository, developing a CLI-based web scraper with Cron scheduling and implementing secure LinkedIn OAuth.
- Integrated the OpenRouter AI provider and refactored the backend architecture to support a secure Bring Your Own Key (BYOK) model, removing deprecated fallbacks.
- Optimized database and API performance using MongoDB compound indexing and RapidAPI rate limiting.
Freelance Web Developer
Self-Employed
- Built and shipped three full-stack projects for small scale startups.
- Scaled the platform to handle 15-20 daily active users by designing secure API endpoints and robust authentication flows, ensuring high availability.
- Digitized local retail operations, increasing overall sales by 50% and expanding customer exposure by 250% within the first month.
Teaching Assistant - Java Programming (BSCS2005)
IIT Madras
- Mentored a cohort of 34 students on core Java concepts, including OOP principles, generics, exception handling, and the Collections Framework.
- Led weekly hands-on troubleshooting sessions and resolved 50+ conceptual queries on the academic discourse forum.
Teaching Assistant - Statistics 1 (BSMA1002)
IIT Madras
- Mentored 150+ students through weekly sessions covering core statistics and probability topics including counting principles, random variables, descriptive statistics, probability fundamentals, hypothesis testing.
Selected Work
Nodeflow
Full-Stack Workflow Automation Platform
Built a custom DAG execution engine using topological sorting to resolve dependencies. Engineered real-time data interpolation and type-safe integrations for diverse APIs alongside robust multi-provider authentication. Integrated OpenAI, Gemini, Stripe, Discord, Forms, Slack.
Throwing around words like "topological sorting" so clients think I'm too smart to argue with.
LinkVault
Centralized Knowledge Canvas & Link Aggregator
Acts as a centralized hub / link manager to organize your links into private or public folders such as ("Books to read", "Next.js Go To Docs", "AI Tools") and it has a deep-cloning mechanism for community-shared collections.
Built this to organize links because I'm physically incapable of closing any of my 143 open browser tabs.
Gradely
Interactive Quiz Application (Academics Proj)
Built a full-stack quiz platform with secure APIs and RBAC. Improved performance and scalability using Redis caching, Celery background jobs, and rate limiting. Developed analytics dashboards for real-time insights.
Built this just for the sake of getting grades.
Comment Category Prediction
Supervised NLP Machine Learning Project (Only Scikit-Learn)
End-to-end pipeline predicting comment category. Heavy EDA, feature engineering, and cross-model comparison with hyperparameter search. TF-IDF + LogReg Baseline, Ensembled LightGBM and LogReg (soft voting).
Restriction on libraries was a pain, but that just made it more fun... right?
Messy Mashup Audio Classification
Audio Classification using Deep Learning Project
Robust music genre classification under heavily degraded, noisy, real-world mashup conditions. Introduced Robust and aggressive Preprocessing and Experimented with Scratch deep CNN (3.2M params), ResNet and EfficientNet-B3 models
Forced the model to listen to audio so bad it practically begged for early stopping... :(
Technical Arsenal
A continuous flow of technologies I use to build scalable platforms and intelligent systems.
- TypeScript
- React
- Next.js
- Tailwind CSS
- Node.js
- Express
- tRPC
- PostgreSQL
- Prisma
- Docker
- Git
- Python
- Flask
- FastAPI
- PyTorch
- Scikit-learn
- NumPy
- Pandas
- Matplotlib
- Seaborn
- SQLAlchemy
- SQLite
- JavaScript
- Java
- SQL
- Redux
- Recoil
- HTML5
- Chart.js
- REST APIs
- MongoDB
- Redis
- GitHub
Let's Connect
I'm always open to discussing new projects, creative ideas, or opportunities to be part of your visions.
© 2026 Kaushal Vaid. All rights reserved.