Sanatan Khemariya

Supercharging legacy AI models with multimodal intelligence—RAG, image processing, code execution, sentiment analysis, and web search—all in a lightweight package.

• Image Analysis with GPT-4o-mini vision for deep visual insights.
• Intelligent Code Execution for real-time debugging.
• RAG-powered Document Chat for PDFs, Word files, and more.
• Web Integration for real-time, dynamic information retrieval with chat context.
• Chat History Retention & Database Integration for conversation memory.
• Multilingual & Speech Support for global accessibility.
• Adaptive Personality Modes and Sentiment Analysis for customized AI experiences.

An AI-powered system to predict customer churn in the banking sector, achieving 87% accuracy on 50,000+ records and providing real-time insights.

• High Accuracy with 87% model performance on 50,000+ records.
• Real-time Churn Insights via an interactive Streamlit dashboard.
• Accelerated Data Prep cutting feature engineering time by 20%.

A fine-tuned BART model and custom transformer prototype for Indian & UK legal summarization.

• Fine-tuned BART Model for domain-specific legal text.
• Custom Transformer Prototype with 50k BPE vocab and multi-head attention.
• PPO-based RL optimizing ROUGE & BLEU scores.
• Hosted on Hugging Face for easy access & inference.

An AI-powered note-taking platform with predictive writing and real-time translation features.

• Smart Next-Word Prediction with 66% accuracy using an LSTM model.
• High Translation Accuracy at 94.6% with a Transformer-based English-Hindi model.
• Optimized DB & Low Latency queries under 100ms using PostgreSQL and Supabase.