I build things that scale, break them to learn more, and automate anything I get tired of doing twice.
From debugging race conditions at 2AM to tuning 3D printer firmware by hand, I live at the intersection of software, systems, and curiosity. I don’t just write code, I engineer solutions, question assumptions, and obsess over details that actually matter.
A Zero-Knowledge encrypted healthcare vault that leverages AWS-powered AI to digitize physical medical records while ensuring absolute patient data sovereignty.
Tech Stack: Next.js (PWA), AWS (Textract, HealthLake, Bedrock, S3, Lambda), Web Crypto API (AES-256-GCM), FHIR.
- Zero-Knowledge Architecture: Engineered a "Privacy-First" healthcare vault where all data is encrypted client-side using the Web Crypto API; plaintext never touches the server, ensuring absolute patient sovereignty.
- AI-Powered Digitization: Built a serverless pipeline using Amazon Textract and Comprehend Medical to transform messy, handwritten prescriptions into structured, searchable medical entities.
- Clinical Interoperability: Implemented global healthcare standards by structuring data into FHIR (Fast Healthcare Interoperability Resources) format and managing records via AWS HealthLake.
- Emergency "Break-Glass" Protocol: Designed a secure, timed-access bypass for paramedics to view critical vitals (allergies, blood group) during emergencies, backed by a cryptographically verifiable audit trail on DynamoDB.
- Clinical RAG Assistant: Integrated Amazon Bedrock to provide doctors with a Retrieval-Augmented Generation (RAG) co-pilot that synthesizes years of patient history into cited, 30-second clinical summaries.
SpeakSense is a multimodal deep learning project that detects when a user is speaking to a virtual assistant by analyzing both audio and video in real time.
Tech Stack: Python, PyTorch, Real-time Audio-Visual Processing, Deep Learning, Embedded Systems AI
- Enhanced User Experience: Developed a wakeword-free system for virtual assistants using multimodal deep learning, boosting user satisfaction by enabling more natural, fluid conversations.
- Improved Accessibility & Engagement: Increased system responsiveness and accessibility by using real-time audio-visual cues to detect user engagement, leading to broader adoption.
- Enabled On-Device AI: Optimized a deep learning model for real-time, low-power processing, making advanced AI practical for everyday consumer electronics.
- Increased Reliability & Privacy: Reduced false activations and enhanced user privacy by using both audio and visual data to ensure the assistant only listens when directly addressed.
Chrome extension that identifies from which movie or TV series the YouTube Shorts video is from!
Tech Stack: JavaScript, HTML, CSS, Chrome Extension APIs, Google AI Platform
- Enhanced User Experience for YouTube Viewers: Developed a Chrome extension to instantly identify the movie or TV series source of YouTube Shorts, significantly reducing the time and effort users spend manually searching for content origins. This directly addresses a common viewer pain point, increasing their satisfaction and content engagement.
- AI-Powered Content Identification: Leveraged a sophisticated AI model to analyze video metadata (title, description, channel, thumbnail) for high-accuracy source identification. This provided users with immediate and reliable information, including specific season and episode details for TV series.
- Streamlined Information Discovery: Created a user-friendly tool that seamlessly integrates into the YouTube interface, allowing for quick and intuitive identification of content. This improved the overall viewing experience by providing valuable context on-demand.
This is AIRAv4 at its initial stage, It has Wave, Hi, Salute and is powered by llama3, gemini and groq.
Tech Stack: C++, Llama 3, Gemini, Groq, AI Integration, CAD Design, Hardware Integration
- Enhanced User Interaction: Engineered an interactive AI agent, project-AIRA, to provide a more natural and engaging user experience. By leveraging advanced AI, the project moved beyond traditional interfaces to boost user satisfaction.
- State-of-the-Art AI Integration: Integrated multiple cutting-edge Large Language Models (LLMs), including Llama 3 and Gemini, to power sophisticated and fluid conversational abilities. This allowed the agent to provide intelligent, real-time responses, significantly improving the quality of user interaction.
- Foundation for Immersive Experience: Developed the core framework in C++, including initial interactive gestures. This foundational work paved the way for a more physically and emotionally resonant AI, creating a more immersive experience for users.
- ESP32 SPIFFS File Upload Tool using Arduino IDE 2x
- Instantly Identify TV Shows & Movies from YouTube Shorts with SceneSniffer Chrome Extension
- Discover Printcalc.me: Your Go-To Tool for 3D Printing Cost Calculations
- I made an autonomous CyberTruck with a Raspberry Pi
- Is Natural Language Processing relevant?