Building Intelligent Systems with AI & Machine Learning
AI Engineer focused on Machine Learning, Deep Learning, LLMs, Computer Vision, and Generative AI. I build things that work in production, not just in notebooks.
From curiosity to intelligent systems
The story behind the work - how I got into AI, what drives me, and where I'm headed.
I'm a B.Tech Computer Science student at Rai University, Ahmedabad, with a focus on AI and machine learning. I spend most of my time building projects that actually run in production, not just notebooks.
My interest in ML started when I wanted to understand how machines pick up on patterns that humans miss. That pulled me into Python, linear algebra, and frameworks like Scikit-Learn, TensorFlow, and PyTorch. I've since built classifiers, neural networks, and LLM-powered apps, and I care deeply about whether a model works in the real world, not just on a test set.
Right now I'm focused on LLMs, Retrieval-Augmented Generation, and applied deep learning. I'm looking for an AI/ML internship where I can work on real production systems alongside people who care about the craft.
Why ML
I want to understand how machines find patterns that humans can't. The engineering required to make those patterns useful is what keeps me going.
Research interests
LLMs, Retrieval-Augmented Generation, applied deep learning, and the gap between a good offline metric and actual user impact.
Career goal
An AI/ML internship where I can contribute to real production systems and learn from engineers who've shipped at scale.
My journey into AI
Started B.Tech in Computer Science
Rai University, Ahmedabad
Began CS degree with a focus on DSA, statistics, and databases. Started building small AI/ML projects on the side to understand how the math actually works in code.
Development Intern Offer
Codveda Technologies
Received an internship offer to build AI/ML and full-stack projects. Learned what it takes to turn an experiment into something that actually ships.
Fell in love with Machine Learning
Self-directed learning
Dove into ML fundamentals: regression, classification, clustering, and ensemble models using Scikit-Learn, NumPy, and Pandas. Built several projects end-to-end and started caring a lot about evaluation, not just accuracy.
Deep Learning clicked
TensorFlow and PyTorch
Trained CNNs for the first time and got hooked. Worked through the engineering side of model training, debugging loss curves, and actual deployment. Built a brain tumor detection app on real MRI data.
Building LLM and GenAI applications
Current focus
Working with LLMs, RAG systems, and AI agents. Trying to combine solid ML foundations with real product engineering - building things people can actually use.
An ML toolkit, end to end
From the math and the models to the frameworks and the deployment - the stack I use to take ideas from notebook to production.
Capability profile
Relative strength across core AI/ML competencies
AI & Machine Learning
Frameworks & Libraries
Programming
Cloud & DevOps
Projects I've built
Real projects - from AI classifiers and NLP models to full-stack web apps. Each ships with a live demo, source code, and a full case study.
Machine learning models, NLP classifiers, and AI-powered applications - trained, evaluated, and deployed.
Generative AI · Full-Stack AIA comprehensive AI workspace featuring 6+ specialized modules powered by Groq.
Deep Learning · CNNAI-Powered Brain Tumor Detection System using a custom Convolutional Neural Network.
Generative AIAI-powered technical assistant for code generation and debugging.
NLP · Machine LearningAI-powered resume analyser using XGBoost + TF-IDF to classify job categories instantly.
NLP · Classification99.29% accurate fake news classifier using Linear SVM on 44K articles.
Regression · MLXGBoost regression model predicting California house prices via a real-time Next.js UI.
ANN · CNN · RNNAI-powered application that generates descriptive captions for uploaded images.
Evaluation metrics, end to end
Every model is only as good as its evaluation. Here are the headline metrics for my trained models - accuracy, precision, recall, F1, and the ROC profile.
ROC profile · AUC 0.940
Brain Tumor CNN
Deep Learning · 4 classes · TensorFlow
Area Under ROC
0.940
Excellent discrimination
Try my models - live
Run real inference against three of my trained models. Upload a file, paste text, and watch the pipeline execute step by step.
Quick examples
Paste text and click Run Inference
How my stack connects
An interactive map of the tools I work with and how they relate - from Python and PyTorch to LLMs, vector stores, and cloud. Hover a node to trace its connections.
← Scroll to explore the full graph →
Always learning, always verifying
Workshops and trainings from my LinkedIn profile - each one verifiable with a direct link.
Industry validation and opportunities
A collection of internship offers I've received, validating my skills in AI, Machine Learning, and Full-Stack Development.
I build in public
A snapshot of my GitHub activity - contributions, top languages, and streaks. Code is the resume that can't be faked.
Contribution graph
@HarshPariyaWhat people say
Recommendations from mentors, teammates, and collaborators.
"Harsh builds things end to end. He doesn't just run experiments in notebooks - he gets the model into a UI and deploys it. That kind of follow-through is rare at this stage."
Let's build something intelligent
Open to AI/ML internships and research roles. If you have a project, a role, or just want to talk ML - reach out.