Portfolio - AMAL NATH VJ

Machine Learning Project

LOG Classifier – LLM‑Powered Log Message Categorization

Domain/Function: Log Analysis, DevOps Monitoring, AIOps, Observability

An LLM-powered Python tool that reads application log messages and automatically classifies them into categories like workflow errors or deprecation warnings. It helps teams triage issues faster and keep monitoring dashboards clean.

View Project
on GitHUB

LOG Classifier – LLM‑Powered Log Message Categorization

Workflow of my project

Project Details

LOG Classifier is a Python-based tool that uses a Large Language Model (via the Groq API) to intelligently categorize raw application log messages. Instead of manually scanning logs, teams can send each log line to the classifier, which returns a clean label such as “Workflow Error”, “Deprecation Warning”, or “Unclassified”.
The project uses environment-based configuration (python-dotenv), prompt engineering, and regular expressions to reliably extract the model’s decision from its response. This makes it easy to integrate into existing monitoring pipelines, dashboards, or alerting systems. The goal is to reduce noise in logs, speed up incident triage, and provide a simple starting point for AI-assisted observability in real-world systems.

Machine Learning Project

Health Insurance Cost Predictor – Age-Aware ML Models

Domain/Function: Health Insurance, Cost Prediction, Risk Assessment

A Streamlit-based machine learning app that predicts individual health insurance costs using demographic and health-related features. It uses separate models for users above and below 25, improving accuracy for younger users by incorporating an additional genetical risk feature.

View Project
on GitHUB

Health Insurance Cost Predictor – Age-Aware ML Models

work flow

workflow

Project Details

Health Insurance Cost Predictor is a data science project that estimates health insurance charges using machine learning models exposed through a Streamlit web interface. The workflow starts with data preparation and feature engineering in Jupyter notebooks, where encoding, cleaning, and exploration are performed.
To handle different error patterns, the project trains two separate models: one for users older than 25 and another for users 25 or younger, where an additional genetical_risk feature is introduced. Both models are saved as artifacts and loaded by the Streamlit app (in app/main.py). At prediction time, the app automatically selects the appropriate model based on the user’s age, providing a more accurate and personalized cost estimate. This project demonstrates end‑to‑end ML: from experimentation and model training to deployment as an interactive web app.

GenAI Project

AskMyDocs – RAG‑Powered PDF Question Answering

Domain/Function: Document Q&A, Knowledge Retrieval, RAG (Retrieval-Augmented Generation)

AskMyDocs is a Retrieval-Augmented Generation (RAG) system that lets you ask natural language questions about your PDF documents and get accurate, context-aware answers. It processes PDFs into vector embeddings using sentence transformers and uses LLMs via GROQ/OpenAI to generate answers with clear

View Project
on GitHUB

AskMyDocs – RAG‑Powered PDF Question Answering

workflow

Project Details

AskMyDocs is an intelligent document question-answering system built on Retrieval-Augmented Generation (RAG). It loads one or more PDF files, extracts their text, and splits the content into smart chunks with configurable size and overlap. These chunks are converted into vector embeddings using sentence transformers and stored in ChromaDB for fast semantic search.
At query time, the system retrieves the most relevant chunks and uses a Large Language Model (GROQ’s Llama 3 or other supported LLMs) to generate a concise, context-aware answer. Each response includes citations back to the original PDFs and page numbers, improving transparency and trust. The project is modular, with separate components for ingestion, index building, and Q&A, and supports environment-based configuration via a .env file. It demonstrates a complete, production-style RAG pipeline for turning static PDFs into an interactive knowledge assistant.

Python Project

Grocery Store Management System – Flask & MySQL Backend

Domain/Function: Retail, Inventory Management, Order Management

A Flask-based backend API for managing grocery store operations, including products, units of measurement, and customer orders. It exposes clean REST endpoints over a MySQL database so a web or mobile frontend can easily handle product catalogs, pricing, and order workflows.

View Project
on GitHUB

Other Tools

python

flask

mysql

Postman

curl

Grocery Store Management System – Flask & MySQL Backend

workflow

Project Details

The Grocery Store Management System is a backend API built with Python Flask and MySQL to support core grocery store operations. It provides endpoints for creating, reading, updating, and deleting products, managing units of measurement (kg, liter, piece, etc.), and handling customer orders. The database is modeled with proper relationships, and all operations go through dedicated DAO modules for products, UOM, and orders.
The Flask server exposes RESTful routes that can be consumed by any frontend, such as a web-based admin UI or a POS system. CORS support makes integration with browser-based clients straightforward. This project demonstrates how to design a small but realistic backend: setting up database connections, defining schema, organizing data access layers, and exposing a clean API surface for a grocery business.

Machine Learning Project

Fake Job Posting Detection System – NLP & ML Web App

Domain/Function: Fraud Detection, HR Tech, Job Posting Screening

A Flask-based machine learning web application that analyzes job postings and predicts whether they are legitimate or fraudulent. It combines TF‑IDF text features, encoded categorical variables, and a trained ML model to help users quickly screen suspicious job ads.

View Project
on GitHUB

Other Tools

python

flask

sql

TF‑IDF

nlp

SciPy

Numpy

Scikit-learn

html

css

git

Fake Job Posting Detection System – NLP & ML Web App

model training

Project Details

The Fake Job Posting Detection System is an end‑to‑end machine learning web app that helps identify potentially fraudulent job postings. The backend is built with Flask and loads a pre‑trained Scikit‑learn model along with multiple TF‑IDF vectorizers for text fields such as company profile, job description, requirements, and benefits. Additional features like telecommuting, company logo presence, screening questions, employment type, experience, education, location (country, city), industry, and function are encoded and fed into the model.
Users interact with a simple HTML/CSS frontend where they enter job details and submit the form. The server preprocesses the input, transforms it using the stored vectorizers and label encoders, and returns a prediction indicating whether the job posting is likely legitimate or fraudulent. This project demonstrates practical NLP, model deployment, and turning a notebook‑trained classifier into a usable web tool.

GenAI Project

LinkLens – AI-Powered URL Content Analyzer

Domain/Function: Web Content Analysis, Knowledge Retrieval, URL Intelligence Other Tools*

LinkLens is a Streamlit web app that analyzes content from multiple URLs and lets users ask natural language questions about those pages. It uses LangChain, OpenAI, and a FAISS vector store to perform semantic search over web content and returns answers with clear source citations.

View Project
on GitHUB

Other Tools

python

ai

langchain

fais

OpenAI GPT

Streamlit

git

vscode

LinkLens – AI-Powered URL Content Analyzer

the tools

Project Details

LinkLens is an AI-powered URL content analyzer and question-answering system built with Streamlit and LangChain. Users provide up to three URLs, which are fetched via Unstructured loaders and split into manageable text chunks using RecursiveCharacterTextSplitter. These chunks are embedded using OpenAI models and stored in a FAISS vector database for efficient semantic search.
When a user submits a question, LinkLens retrieves the most relevant chunks and uses a retrieval-based QA chain to generate an answer along with source citations pointing back to the original URLs. Environment variables (via python-dotenv) are used to securely manage the OpenAI API key. The app offers a clean, real-time interface that makes it easy to inspect and interrogate web content without manually reading each page.

Hello, I am

AMAL NATH VJ

3

2

1

About

Key

My

LOG Classifier – LLM‑Powered Log Message Categorization

Health Insurance Cost Predictor – Age-Aware ML Models

AskMyDocs – RAG‑Powered PDF Question Answering

Grocery Store Management System – Flask & MySQL Backend

Fake Job Posting Detection System – NLP & ML Web App

LinkLens – AI-Powered URL Content Analyzer

My

Experience

Software Developer – UVJ Technology, Trivandrum

Intern – Siemens Technology and Services Pvt. Ltd., Chennai

Awards

Contact Me

Email

Linkedin

AMAL NATH VJ

Download Resume