Project – Rauhan Ahmed

Chatting with Champions: An AI Conversational Football Historian

Want to grill a legend like Cristiano Ronaldo or Lionel Messi with questions? This project builds an AI that lets you chat with the 35 greatest footballers EVER! Ask anything about their careers, rivalries, or iconic moments, and this AI, powered by cutting-edge tech, will answer like a true football mastermind. Get ready to rewrite fan history!

VIEW PROJECT

Overview

This project delves into the world of football legends, offering a unique and interactive way to explore their careers and achievements. By leveraging cutting-edge AI techniques, it empowers users to ask insightful questions about any of the top 35 footballers of all time, as identified by a renowned source like The Guardian.

Tech Stack Used

Here’s a breakdown of the system’s core elements:

Data Foundation: The project starts by building a solid data foundation. It gathers information on the 35 greatest players through web scraping techniques, specifically targeting their Wikipedia pages. This rich textual data becomes the fuel for the AI engine.
RAG System Construction: A Retrieval Augmented Generation (RAG) system is then constructed. This innovative technology bridges the gap between traditional information retrieval and powerful language models. Simply put, the RAG system guides the AI towards relevant information within the collected data, enabling it to provide more focused and accurate responses.
OpenAI GPT-3.5 Turbo Power: As the engine for generating responses, the project utilizes OpenAI’s GPT-3.5 Turbo model. This state-of-the-art language model possesses remarkable capabilities in understanding and responding to natural language inquiries.
LangChain and Embeddings: To further enhance the system’s performance, langchain, a powerful framework, is employed. Additionally, free, open-source embeddings from Hugging Face are integrated. These embeddings essentially translate text into a numerical format, allowing the AI model to grasp the relationships between words and concepts with greater precision.
FAISS and Local Vectorstore: For efficient information retrieval, FAISS, a library for fast approximate nearest neighbors search, is utilized. This allows the system to quickly locate relevant information within the player data based on the user’s query.
Gradio Interface: Finally, a user-friendly Gradio interface is built. This interface acts as the bridge between the user and the AI system. Users can simply type their questions about any of the 35 legendary footballers, and the system will generate insightful and informative responses, drawing upon the vast knowledge base it has constructed.

In essence, this project offers a captivating example of how advanced AI techniques can be harnessed to create a truly engaging and informative experience. Football enthusiasts can now delve deeper into the stories of their favorite players through an interactive and intelligent interface, fostering a deeper appreciation for the sport’s rich history.

Project

Conversational Data Analysis with Groq Inference Engine and Mixtral-8x7B

Talk data, get insights. This project uses Groq’s LPU for lightning-fast conversational analysis on a massive 8x7B LLM. Upload CSV/Excel, ask questions, and see charts & insights – all in a sleek Streamlit interface on HuggingFace Spaces. Data analysis redefined.

VIEW PROJECT

Overview

This project pushes the boundaries of data exploration by creating a web application that seamlessly blends natural language with powerful data analysis. At its core lies the innovative Groq LPU Inference Engine, the world’s first built specifically for large language models (LLMs). This groundbreaking technology utilizes a custom-designed Language Processing Unit (LPU) to deliver exceptional performance and precision.

Here’s the magic: you simply upload a CSV or Excel file, and the application leverages the mighty Mixtral 8x7B LLM to understand your questions in natural language. PandasAI, a master of conversational data analysis, then jumps in to perform the analysis you requested. The results are presented in a user-friendly Streamlit interface, complete with charts and summaries, all accessible through a web browser.

This innovative combination allows you to ditch the complex world of code and focus on uncovering insights from your data using natural language.

Tech Stack Used

This project brings together a powerhouse of technologies to deliver an exceptional user experience

Groq LPU Inference Engine: Used for powering the Mixtral 8x7B LLM, enabling exceptional sequential performance, high accuracy, and instant memory access for natural language processing tasks.
Mixtral 8x7B LLM: A large language model from Groq, providing the core functionality of understanding and responding to user queries from uploaded data files.
PandasAI Library: Facilitates conversational data analysis, allowing the LLM to interpret and analyze data from CSV/Excel files in natural language.
Streamlit: Used to create the web application interface, enabling users to interact with the LLM and visualize results through charts and summaries.
Hugging Face Spaces: Platform for deploying the web application, making it accessible to users online.

This synergistic blend of technologies empowers you to unlock the secrets of your data with unprecedented ease.

Project

212

Multiple-Multivariate Time Series Forecasting

End-to-End Python based ML project focusing on forecasting multiple multivariate time series with production grade deployment techniques.

212

VIEW PROJECT

Overview

This is a Python, ML-based API with an impressive Normalised Mean Squared Error of 0.039 and Normalised Mean Absolute Error of 0.10 for forecasts of this multiple, multivariate timeseries data on a 15-day window. This POC aims to empowers retailers with data-driven sales forecasting to optimize inventory management and improve profitability. This comprehensive solution predicts short-term sales (up to 30 days) with outstanding accuracy and performance metrics. By integrating seamlessly with existing workflows, it enables retailers to:

Make informed inventory decisions.
Reduce stockouts and overstocking.
Improve operational efficiency.
Boost customer satisfaction.

Tech Stack

Darts : For efficient time series operations and forecasting.
MongoDB : For storage and retrieval of data.
LightGBM : To accurately predict covariate and target features.
Scikit-learn : For creating data pipelines.
DVC, Git, and Github : For seamless data and code versioning.
Evidently AI: To check for data drift/target drift.
FastAPI : For building a user-friendly API for model accessibility.
Docker and Dockerhub : For secure and streamlined deployment to the Render Cloud Platform.
Github Actions : For automating the CI/CD pipeline.
Render Cloud : A PaaS for deployment of web apps, api’s, etc.

Project

HyperRez: Sharpen Blurry Memories

HyperRez: An open-source python library to make your blurry photos shine in just 2 lines of code!

VIEW PROJECT

Overview

Enhance and upscale your images effortlessly with HyperRez, a Python library crafted by Rauhan Ahmed Siddiqui. Leveraging the power of GFPGAN and RealESRGAN, state-of-the-art Generative Adversarial Networks, this library provides advanced image enhancement capabilities. Real-ESRGAN focuses on upscaling backgrounds, while GFPGAN excels in refining the faces of human subjects.

Features

Dual Network Integration: Combines Real-ESRGAN and GFPGAN for comprehensive image enhancement.
Gradio Interface: Seamlessly deployed into production using Gradio Interface for an interactive and user-friendly experience.
GPU Acceleration: Optimized for GPU processing, ensuring fast and efficient image enhancement (Note: Performance may be suboptimal on CPUs).
PyPI Integration: Available on PyPI as the “HyperRez” library for easy installation and use.

Project

Term Deposit Prediction

A comprehensive end-to-end Machine Learning project designed to predict bank deposit subscriptions using the well-known “Bank Marketing” dataset with production grade deployment techniques.

VIEW PROJECT

Overview

This project tackles the challenge of predicting bank deposit subscriptions using a comprehensive Machine Learning pipeline. It leverages the popular “Bank Marketing” dataset to train and deploy a model that optimizes F1-score and important performance metrics due to the dataset’s imbalanced nature.

The project demonstrates a complete workflow, including data ingestion with CockroachDB, data exploration and feature engineering, model selection with hyperparameter tuning via MLFlow, and deployment to Microsoft Azure cloud using Docker containers and Github Actions for CI/CD. Additionally, a web application is built with PyWebIO and Flask for user interaction with the prediction model.

Tech Stack Used

Data Storage & Retrieval

CockroachDB (SQL Database): Efficiently stored and retrieved the “Bank Marketing” dataset for data ingestion..

Data Preprocessing & Analysis

Python Libraries (Pandas, NumPy, Seaborn, etc.): Handled data manipulation, cleaning, exploration, and visualization, providing insights for feature engineering.

Model Selection & Model Training

Machine Learning Algorithms (CatBoost, Random Forests, SVM, etc): Built, trained, and evaluated various machine learning models to predict bank deposit subscriptions.
MLFlow: Tracked model experiments, hyperparameter tuning results, and other metrics (AUC-ROC, Precision, Recall) for optimal model selection.
DagsHub: Remote server for comparing different experiments tracked by MLFlow.

Web Application Development

Docker: Containerized the prediction pipeline, ensuring consistent runtime environment across deployment stages.
PyWebIO & Flask: Developed an interactive web interface for users to leverage the trained model’s predictions.

Model Deployment

Microsoft Container Registry (MCR): Securely stored the Docker image for deployment.
Microsoft Azure Web App Service: Hosted the web application built for user interaction with the prediction model.

CI/CD (Continuous Integration/Continuous Delivery)

GitHub Actions: Automated the build, testing, and deployment pipeline for efficient and reliable updates.