I build AI that listens, sees, and answers.

Report and Interview Integration Client work

AI platform that turns interview recordings and a draft into a structured assessment report. It transcribes the audio, merges the transcripts into the draft section by section, and writes an analysis of the interview.

Tools: LLM, Speech-to-Text, Vertex AI, Gemini, DeepSeek, FastAPI, Celery, PostgreSQL, Gotenberg, Docker

Research Assistant Bot, source on GitHub

Research Assistant Bot (RAB) is a user-friendly research tool for retrieving information. Users can input article URLs or upload a PDF file and ask questions related to the content.

Tools: GenAI, LLM, LangChain, Groq, FAISS, Llama 3, Streamlit

Outreach Mail Generator, source on GitHub

Outreach Mail Generator is a tool designed to help companies generate outreach emails using Groq, LangChain, ChromaDB, Llama-3.3-70B-Versatile, and Streamlit/Gradio. Users can input the URL of a company's careers page, extract job postings, and generate personalized emails. These emails include relevant portfolio links retrieved from a vector database, tailored to match specific job descriptions.

Tools: GenAI, LLM, LangChain, Groq, ChromaDB, Llama 3, Streamlit, Gradio

Personal Medical Chatbot with LLM RAG, source on GitHub

Personal medical chatbot project using Large Language Models (LLM) and Retrieval Augmented Generation (RAG). The chatbot retrieves medical information from PDF documents, employs LlamaCPP for LLM integration, and uses llama_index for efficient document handling. This approach enhances natural language responses and advanced information retrieval.

Tools: GenAI, LLM, RAG, LlamaCPP, LlamaIndex, Gradio

Simple Personal Chatbot, source on GitHub

This simple personal chatbot project uses Artificial Neural Network (ANN) to process user queries, utilizing Flask and TensorFlow for development. It performs data preprocessing by tokenizing, removing punctuation and stopwords.

Tools: Rule-based, NLP, TensorFlow, Keras, Flask, ANN

Plant Disease Detection, source on GitHub

This plant disease detection system uses traditional ML methods such as SVM and Random Forest with feature extraction techniques (Hu Moments, Haralick textures, and Color histograms) and pretrained CNNs such as ResNet50V2 and MobileNetV2. It includes a FastAPI backend with TensorFlow Serving. The model is deployed on Google Cloud Platform (GCP) using Cloud Functions and Cloud Storage, with an additional Gradio interface for easy access on Hugging Face Spaces.

Tools: TensorFlow, Keras, ResNet, MobileNet, TF Serving, FastAPI, Google Cloud Platform (GCP), Google Cloud Storage, Google Cloud Functions, Scikit-learn, Gradio

Speech Emotion Recognition, source on GitHub

Speech Emotion Recognition (SER) using a Convolutional Neural Network (CNN). The project utilized datasets such as CREMA, RAVDESS, SAVEE, and TESS to train the model to classify various emotions from audio recordings.

Tools: TensorFlow, Keras, CNN

LibraLink: Smart Attendance System with Book Recommendation, source on GitHub

LibraLink: An advanced library system integrating facial recognition for attendance, using CNN or LBPH algorithms based on facial data patterns. It also features a book recommendation service, matching user preferences through the K-Nearest Neighbors (KNN) algorithm. The system uses MySQL as its relational database to manage and store student data.

Tools: TensorFlow, OpenCV, CNN, LBPH, KNN, Tkinter, MySQL

Image Classification of Billionaires, source on GitHub

This project contains a classification of billionaires, namely 20 billionaires.

Tools: SVM, OpenCV, Flask

Credit Card Fraud Detection Model, source on GitHub

Machine learning model to detect credit card fraud using various algorithms such as Logistic Regression, Decision Tree, Random Forest, and XGBoost. This dataset is highly imbalanced, so several resampling techniques are applied to balance the data.

Tools: Scikit-learn, XGBoost, Resampling Techniques

Credit Risk Model, source on GitHub

Credit risk prediction model aimed at predicting customers who are at potential risk of defaulting on their loans based on historical loan data.

Tools: Scikit-learn, MLflow

Patient’s Condition Classification Using Drug Reviews, source on GitHub

Patient condition classification using drug reviews applies multiple machine learning models such as Passive Aggressive Classifier, Random Forest, XGBoost, and etc. The project also applies text processing techniques including CountVectorizer, TFIDFVectorizer, and Word2Vec to analyze user reviews from a dataset. The project uses Streamlit for the user interface and deployment to provide a demo.

Tools: NLP, Scikit-learn, XGBoost, Word2Vec, Streamlit

Twitter Sentiment Analysis, source on GitHub

This project is a sentiment analysis of Twitter, where I use deep learning models like MLP and LSTM to classify tweets based on positive or negative sentiments.

Tools: NLP, TensorFlow, Keras, LSTM

PySpark ML, source on GitHub

This PySpark_ML repository focuses on data processing and machine learning, covering PySpark concepts such as data manipulation, transformation, and aggregation. It includes machine learning techniques using Spark's MLlib and ML libraries, with implementations of various regression and classification algorithms (e.g., random forest regressor, linear regression, logistic regression, naïve bayes, gradient boosting, and random forest classifier).

Tools: PySpark, MLlib, ML

Movie Recommendation System, source on GitHub

This ML project utilizes a Content-Based Recommender System for movies, using features like titles, genres, and posters to provide recommendations based on the chosen film.

Tools: Recommender System, Content-Based, Cosine Similarity, Streamlit

Movie Recommendation System (Coursera course), source on GitHub

These are assignments from a course in the Coursera Machine Learning Specialization, which involve implementing collaborative filtering and content-based filtering recommender systems using techniques such as matrix factorization and neural networks.

Tools: Recommender System, Content-Based, Collaborative Filtering, TensorFlow, Matrix Factorization

Ad Hoc Insights: Consumer Goods, source on GitHub

Analyzed consumer goods data using MySQL with Joins, Subqueries, Functions, CTEs, etc. Insights were visualized in Tableau and presented using Canva.

Tools: MySQL, Tableau, Canva, Ad Hoc Analysis

Food and Beverages OOP Concepts, source on GitHub

OOPs concept using Java and connect to MySQL Database in Food and Beverage (FnB).

Tools: Java, JDBC, MySQL, OOPs, JavaFX