Mayank Mittal

A distributed Kerberos-like authentication system using 2-of-3 Schnorr multi-signatures that remains secure even when one authority is fully compromised — no single point of failure.

CipherRounds is a stateful symmetric-key protocol enabling secure multi-client data aggregation in hostile networks using AES-CBC encryption, HMAC authentication, and cryptographic ratcheting to prevent replay, tampering, and desynchronization attacks.

CipherFleet is a cryptographically secure UAV Command-and-Control system implementing manual ElGamal cryptography, mutual authentication, session key establishment, and AES-256 encrypted fleet communication with replay and MITM attack defenses.

MediMind: RAG-powered medical chatbot leveraging LangChain, OpenAI GPT, and Pinecone vector DB for semantic retrieval. Flask-served UI with HuggingFace embeddings (all-MiniLM-L6-v2) enables context-aware medical query responses from PDF knowledge bases.

PrediFly: Cutting-edge drone trajectory prediction merging YOLOv8 detection, Kalman filtering, and GRU-CNN neural nets for high-precision aerial tracking across complex flight patterns—computer vision meets deep learning.

🤖 Rezzy – A Full-Stack, LLM-Integrated AI Study Assistant for Medical Students Featuring Flashcard & Question Retrieval via Semantic Vector Search, Embedding-Based Similarity, Retrieval-Augmented Generation (RAG), and Real-Time Response Streaming with WebSockets & Server-Sent Events 🧠📚⚡

Web app for geospatial inference via image—React + FastAPI frontend/backend, ConvNeXt, Swin Transformer, EfficientNet models for region, lat-long, and angle prediction; GAN-augmented training, real-time multi-model inference, gradient-optimized.

CalcuTron is a transformer-based sequence-to-sequence model engineered for symbolic arithmetic reasoning. Leveraging multi-head self-attention, positional encoding, and deep encoder-decoder layers, it learns to perform multi-digit addition and subtraction. It generalizes to longer sequences without explicit rules, showcasing emergent algorithmics.

Chatterly: A high-performance Socket.io-powered real-time messaging platform with JWT-secured microservices architecture, React/Zustand reactivity, MongoDB persistence layer, and TailwindCSS/DaisyUI theming—delivering blazing-fast WebSocket communication with zero compromise on security.

Interactive website to maintain the hospital warehouse and keeping record of all important things like medicines,equipments etc and also the log of patients

ByteEats 🍽️🤖 is an AI-driven food ordering chatbot using FastAPI, SQLite3, and Dialogflow, with Ngrok/Cloudflare Tunnel for accessibility and a sleek HTML/CSS frontend. It enables seamless order management, tracking, and modifications in real time. 🚀🔥

No description provided.

An end-to-end Age Prediction system leveraging Deep Learning, combining a PyTorch-based Custom CNN and fine-tuned ResNet-18 models, seamlessly integrated with a FastAPI backend for asynchronous inference and a React.js + TailwindCSS frontend delivering an intuitive and responsive user interface.

PageSage is a modular web scraping and semantic retrieval system using BeautifulSoup for extraction, SentenceTransformers for dense embeddings, and ChromaDB for scalable vector search—enabling efficient data indexing for advanced RAG pipelines.

Detecting duplicate Quora question pairs using ML, deep learning (LSTM, BERT), advanced NLP, feature engineering, and ensemble methods—optimized for accuracy and scalability.

This repository contains implementations of various machine learning algorithms from scratch, including Multi-Layer Perceptron (MLP), Gaussian Mixture Models (GMM), Principal Component Analysis (PCA), Autoencoders, and Variational Autoencoders.

Implementation of various machine learning algorithms from scratch, including Linear Regression, K-Nearest Neighbors, Decision Trees, and K-Means clustering. Also done EDA on data, Implemented LSH, IVF, SLIC algorithms also with evaluation metrics

No description provided.

Contains slides, notes, assignments ,pyq and project works of all semesters

Powerful neural architectures for next word prediction using FFNN, RNN, and LSTM models

An implementation of n-gram language models with various smoothing techniques for natural text generation and analysis with tokenization and perplexity calculations

An implementation of static word embedding techniques for natural language processing, featuring SVD, CBOW, and Skip-gram models.

A computational physics project exploring critical phenomena through percolation theory and the Bak-Tang-Wiesenfeld sandpile model.

This computational chemistry project delves into quantum mechanical analyses of various molecular systems using state-of-the-art computational methods. Through the application of Hartree-Fock (HF) theory with the STO-3G basis set in Gaussian software, we explore reaction energetics, molecular conformations, and spectroscopic properties

Code base to basically decrease the waiting time at the canteens across campus without any delivery personnel. We tend to order the food and do the online payment and then also giving the facility to track the order and also doing the concurrency of orders at canteen level and also providing the live food items available to save time of students..

MiniGFS is a simplified Google File System (GFS) replica, implemented in Python. It simulates a distributed file storage system using one master server, chunk servers, and a client. The system handles file uploads and reads, distributing data into 64-byte chunks and managing metadata.

GeoShield is an advanced AI-driven platform designed for real-time landslide risk assessment using interactive satellite imagery and deep learning

This repository is for the frontend and backend code of the Gait and Behaviour Based Autism Screening Tool as part of the Design for Social Innovation course project.

Stock price predictor LSTM Sequential Model with Dropout Regularization by which we can analyse any stock tickers, do its fundamental analysis using fundamental ratios and charts visualisations of 100MA and 200MA and can also predict stock price for next 10 days with its trend. Can also view candle stick charts for stock trading and latest news.

This is a web application designed to assist doctors in managing the treatment of several patients belonging to the 25 ambulances alloted to him using data such as patient vitals, accessing medical documents, and communicating with Emergency Medical Technicians (EMTs) for critical cases.