AI & ML Solutions Portfolio
Discover our AI/ML portfolio, featuring cutting-edge artificial intelligence and machine learning solutions designed to unlock data-driven insights and optimize business performance. Explore how our AI and ML technologies can transform your business and drive success.
Implicit Feedback Reccomendor
This project builds a hybrid recommendation system for personalized news using the MovieLens dataset as a proxy. It models implicit feedback (treating ratings ≥ 4.0 as positive interactions) and incorporates content metadata (genres and tags) to generate Top-N recommendations using the LightFM library.
Predictive Modeling for Loan Risk Assessment
This project focuses on building a supervised machine learning model to classify loans as either healthy (Class 0) or high-risk (Class 1). The analysis is conducted using a dataset of 77,500 loan entries, incorporating financial attributes such as loan size, interest rate, borrower income, debt-to-income ratio, and credit history metrics.
Heart Disease Predictor
This project predicts the presence of heart disease in a patient using various machine learning models based on clinical features. The dataset used is a standard heart disease dataset containing several medical attributes and a target variable. The notebook explores various classification models including Logistic Regression, Naive Bayes, SVM, KNN, Decision Tree, and Random Forest .
Debt Collector Chatbot
The Debt Collector Chatbot project uses the Llama-3.1-70b-versatile model via the Groq platform to generate intelligent responses for debt-related conversations. It integrates Groq’s API to enable smooth communication between the chatbot and the AI model. The system is designed to assist in tasks like payment reminders, debt negotiations, and schedule handling.
Sales Team Performance Analysis Using LLM
The Sales Team Performance Analysis system uses a Large Language Model via Hugging Face API to provide insights from uploaded sales data in CSV or JSON format. It delivers AI-generated feedback for individual sales reps, team performance summaries, and sales trend forecasting. Built with FastAPI, Uvicorn, Pandas, and Hugging Face Transformers, it offers a powerful backend solution for sales analytics.
Analyzing Sales Performance using BART
This project is a Flask-based backend service integrated with a Hugging Face transformer model (BART) to analyze sales performance data. It's designed for easy deployment (e.g., using Google Colab + ngrok) and provides endpoints to upload sales data and generate AI-powered feedback.The Hugging Face summarization model (facebook/bart-large-cnn) is used to generate executive summaries, personalized insights.
TalkTrackDB
TalkTrackDB is a speech-to-text pipeline that leverages the power of SpeechBrain and Wav2Vec2 to transcribe audio files and store the results in a lightweight SQLite database. Ideal for developers, researchers, and hobbyists working on voice-driven applications or audio archiving.Speech-to-Text Conversion using the Wav2Vec2 model via SpeechBrain SQLite Integration for storing and querying transcriptions
Voice Similarity
To determine whether a given voice recording matches any known speaker in a local database using advanced embedding models for speaker recognition, with data stored in a lightweight SQLite system. Users upload `.wav` files, which are processed using torchaudio, and speaker embeddings are extracted using Wav2Vec2 or ECAPA-TDNN. These embeddings are stored in a SQLite database and compared via cosine similarity to identify or register speakers.