Featured Work
Production-grade AI systems and research tools — built, deployed, and maintained.
End-to-end LLM fine-tuning pipeline with LoRA, QLoRA, and full fine-tuning. Experiment tracking, evaluation metrics, and production-ready model export.
LangGraph-orchestrated framework routing queries to specialized AI agents that autonomously synthesize structured research reports.
Automated dataset curation pipeline filtering, deduplicating, and quality-scoring raw corpora at scale for LLM fine-tuning workflows.
AI-powered review analysis tool that scrapes, classifies, and summarizes customer reviews at scale with sentiment insights.
Autonomous agent that plans multi-step investigations, browses the web, and compiles evidence-backed research reports end-to-end.
Interactive tool for visualizing neural network architectures and training dynamics — watch gradients and activations evolve in real time.
All Projects
The full archive — experiments, explorations, and everything in between.
Agentic content generation platform with RAG, pgvector search, multi-LLM orchestration, and streaming pipelines.
A novel Drug-Target Affinity (DTA) prediction model combining: - **ESM-2** protein language model for deep protein representation - **Graph Attention ...
AI-powered academic paper discovery and management platform. Search ArXiv papers with AI-enhanced relevance scoring, organize them into folders, and create public or private reading lists to share with the research community.
A real-time meeting transcription and AI-powered summarization tool using Google's Gemini API. This application provides speaker diarization, automati...
An agentic RAG pipeline that processes URLs and PDFs concurrently, summarizes each source with Google Gemini AI, and synthesizes the results into a structured academic review paper via a Gradio web interface.
FastAPI-powered Gemini Q&A agent for secure document uploads, OCR, PII redaction, and intelligent search with LangChain & FAISS.
A Python implementation of Snake where neural networks evolve through NEAT algorithm to master the game, showcasing the power of evolutionary AI in learning complex gameplay strategies.
A Flappy Bird clone with AI learning capabilities using the NEAT (NeuroEvolution of Augmenting Topologies) algorithm. Watch as neural networks evolve to master the game through generations of training, or run your best-trained models to see perfect gameplay. Built with Pygame.
A feature-rich Android chess application that offers multiple game modes and a sophisticated chess-playing experience. - *Multiple Game Modes* - Onl...
A classic Pong game implementation for Android devices with modern features and smooth gameplay. This is a 2D Pong game developed for Android that fea...
The Metro Station Planner is a Java application designed to predict, analyze and optimize metro station locations using geographical data. It uses clu...
This project predicts stock prices using advanced Deep Learning models (LSTM-CNN, Bidirectional LSTM, Attention LSTM) and includes a comprehensive Str...
This project implements a **Book Recommendation System** leveraging a dataset containing book details, user information, and ratings. The primary goal...
Face recognition is the ability to look at the digital image of a human and recognize the person just by looking at the face. FaceNet was introduced i...
This project aims to detect and classify malware using various machine learning techniques. It leverages supervised and unsupervised learning algorith...
hey, i'm Yashvardhan Gulyani 👋 building intelligent systems & high‑impact web experiences portfolio •...
View on GitHubAn email spam detector with 99% accuracy made using Naive Bayes classifier with sklearn python library and matplotlib and wordcloud for data represent...