Developed a medical document summarization tool using a secure RAG pipeline with mandatory PHI redaction. The system supports automatic redaction of sensitive fields (name, SSN, MRN) and optional manual redaction based on user-defined values. After redaction, embeddings are generated using a clinical model (MedEmbed), relevant context is retrieved, and summaries are produced using GPT-5.2.
Here is a demo for implementing MCP-based integration for connecting AI assistants with external tools and workflows. The solution demonstrates how structured tool access can enable more reliable task execution, improved automation, and controlled data flow between systems. It was designed to validate integration patterns, support rapid prototyping, and demonstrate real-world AI assistant capabilities in… Continue reading MCP Integration Demo
Developed an EHR-focused chatbot integrated with a Neo4j knowledge graph to deliver accurate, context-aware responses from structured healthcare data. The solution connects patient and clinical relationships into a graph-based model, enabling faster retrieval and improved query understanding. It was designed to support medical workflows, enhance information access, and reduce time spent searching across complex EHR… Continue reading EHR Chatbot & Neo4j Knowledge Graph
Developed a data processing and ETL workflow to clean, transform, and organize raw datasets into structured, analysis-ready outputs. The solution automated key steps such as data validation, formatting, and pipeline execution to improve accuracy and reduce manual effort. It was designed to support scalable reporting, faster decision-making, and reliable data delivery for downstream analytics and… Continue reading Data Processing & ETL Pipeline Automation
Developed a PDF OCR solution to extract readable text from scanned documents and image-based PDFs. The system automates text detection and conversion, enabling users to quickly search, copy, and process document content. It was designed to improve document accessibility, reduce manual data entry, and support efficient extraction for business workflows such as records, forms, and… Continue reading PDF OCR Automation
Developed an image processing model for an automotive company using the U-Net architecture to extract and process key visual details from vehicle images. The solution was deployed and hosted on Hugging Face, enabling easy access, testing, and integration. It was designed to improve automation, reduce manual image review, and support faster processing of automotive inventory… Continue reading Image procceing model
Built and integrated an AI-powered medical chatbot using the OpenAI GPT-3 API to help a healthcare business automate patient conversations and improve online service inquiries. The chatbot handles common questions, guides users toward the right services, and supports lead generation through natural, human-like responses. It was designed to improve customer engagement, reduce manual support workload,… Continue reading Medical Chatbot Integration
Developed an automated voice summary solution that converts text-based content into high-quality audio outputs. The system generates accurate summaries and produces downloadable audio files for easier consumption. It was designed to support faster information delivery, improve accessibility, and enable users to listen to summarized content on demand. The solution integrates text processing, summarization, and text-to-speech… Continue reading Voice summary or audio summary
Developed “Eve,” an AI chatbot designed to provide accurate information and real-time assistance during emergencies. The solution was built using Amazon Bedrock with the Llama 2 Chat 13B foundation model in a serverless architecture. It was designed to deliver fast, reliable responses by sourcing information from verified emergency resources and guiding users through critical situations.
Developed a site selection tool that gathers detailed land parcel data using the Regrid API. The solution supports two search modes: a standard structured search and an AI-powered prompt-based search that interprets user intent and converts it into accurate query parameters. The tool was designed to streamline property research, improve search accuracy, and deliver faster… Continue reading Site selection Tool
