Cultural Appropriateness Bot
Python | BERT | NLP | Streamlit
Developed a BERT-based AI model to classify cultural appropriateness in images and text, complete with a real-time analytics dashboard for monitoring classification trends.
ML (Software) Engineer & AI Enthusiast
Passionate Computer Science graduate leveraging Python, BERT, and cloud systems to transform complex data into intelligent, efficient applications.
Data processing, backend systems, and scripting.
ExpertPandas, NumPy, Scikit-Learn, BERT, and TensorFlow.
ProficientSQL, MongoDB, database design, and data modeling.
ProficientPower BI, Tableau, Streamlit, and Matplotlib.
FamiliarManaging web infrastructure and corporate systems (AWS, etc.).
FamiliarPython | BERT | NLP | Streamlit
Developed a BERT-based AI model to classify cultural appropriateness in images and text, complete with a real-time analytics dashboard for monitoring classification trends.
Python | SQL | MongoDB | Flask | Streamlit
Analyzed user behavior and task efficiency, optimizing task scheduling using reinforcement learning-based analytics.
Python | BERT | Simple Transformers
Designed and implemented a QA system using transformer-based models. Processed and cleaned datasets to improve response accuracy.
2023 - PRESENT
Jacobs Biomedical Company (Freelance)
Architecting and managing the company's complete digital presence, including web infrastructure and corporate communication systems. Ensuring high availability and seamless scalability for business operations.
2023 - 2024
IEEE Chapter @ University of Greater Manchester
Organized events, promoted IEEE initiatives, and engaged with the student community to foster a strong interest in technology, research, and professional development.
If you have a project idea, a technical challenge, or just want to connect about Computer Engineering, drop me a message.
Send a Message NowPython | SQL | MongoDB | Flask | Streamlit
Modern families often struggle with effective task management, especially in multi-child households. Assigning tasks, ensuring completion, and providing appropriate rewards is challenging. This project addresses the difficulty parents face in keeping children motivated and finding a structured, fair way to assign tasks that match their skills and interests.
I developed a full-stack, AI-driven task recommendation system. The solution is a web application (React, Node.js, MongoDB) where parents can manage tasks and teens can receive recommendations. The core of the system is a Reinforcement Learning (RL) model, specifically Proximal Policy Optimization (PPO), which I trained to act as an intelligent recommender. This AI agent suggests the best-suited teen for a task based on priority, skill category, and deadlines, effectively gamifying task management.
Python | BERT | NLP | Streamlit
Mainstream AI models are "culturally blind" and often perpetuate racial and cultural biases. AI image generators create stereotypes, and content moderation tools fail to understand cultural nuances, leading to miscommunication and harmful outputs. This project tackles this problem: the critical lack of culturally aware AI systems, specifically within the Indian cultural context.
I developed a "Cultural Appropriateness Bot" that classifies text as culturally "Appropriate," "Inappropriate," or "Neutral." The core of the solution was building a fine-tuned BERT-based model. Since no public dataset existed for this task, I engineered a high-quality, fully synthetic dataset (named CST02) using GPT-4, focusing entirely on nuanced Indian cultural scenarios. This model achieved a **94% test accuracy**. I also built a complete ecosystem around it: a Flask-based web UI for users to test sentences and a Streamlit dashboard for admins to analyze model performance and user feedback.
Python | BERT | Simple Transformers
There is often a gap between pure entertainment and effective education. This project explores how to bridge that gap by creating an engaging tool that leverages a popular medium (anime) to provide valuable, factual information (science and culture).
I developed a specialized, BERT-based Question-Answering (QA) system designed to engage with anime-themed content. I used the `bert-base-uncased` model and fine-tuned it using the Simple Transformers library. The model was trained on a diverse, custom-built dataset comprising anime transcripts, fan discussions, and related scientific concepts, allowing it to interpret and respond to user queries with high accuracy.