Hi, It's Harsh

Description:

VISUAL-AI is a multimodal AI platform that integrates text, voice, image, video, and gesture-based interactions into a unified intelligent interface. The system combines natural language processing, speech recognition, and computer vision to enable adaptive human-AI communication. It also features an interactive 3D visualization layer that represents AI activities in real time, making complex AI interactions more intuitive and engaging.

Key Features & Achievements:

• Developed a multimodal AI framework integrating 5+ AI models to process text, speech, visual inputs, and gesture-based commands within a single interface.
• Implemented Natural Language Processing, Speech Recognition, and Computer Vision pipelines to enable hands-free interaction and context-aware responses.
• Built a real-time 3D AI visualization interface using Three.js and WebGL, representing system activity and interactions through an interactive globe.
• Designed adaptive response mechanisms to improve contextual understanding and conversational interaction flow.
• Implemented AI-driven content generation capabilities supporting automated text responses and visual outputs.

Technologies Used:

[Python, TensorFlow, NLP, Speech Recognition, Computer Vision, Three.js, WebGL]

Position:

AI Developer

Learnings and Benefits:

Gained experience in multimodal AI system design, human-AI interaction, real-time visualization, and integrating multiple AI models into a unified application architecture.

Description:

CADV is an AI-driven medical imaging system designed to detect cardiac abnormalities from MRI and CT scans using deep learning–based segmentation and visualization. The platform processes volumetric medical data to identify abnormal cardiac structures, generate heatmaps and segmentation masks, and render interactive 3D visualizations. It also provides synchronized multi-view slices (axial, coronal, and sagittal) to help clinicians easily interpret structural abnormalities and improve diagnostic accuracy.

Key Features & Achievements:

• Developed a deep learning segmentation pipeline using an Attention U-Net architecture with residual gates, trained on the CARE-Cardiac dataset for accurate multi-structure cardiac segmentation.
• Implemented automated abnormality detection with heatmaps and segmentation masks to highlight pathological regions within MRI/CT scans.
• Built synchronized multi-axis slice visualization (axial, coronal, sagittal) enabling detailed anatomical inspection and anomaly localization.
• Created an interactive 3D volumetric renderer to visualize cardiac structures and abnormalities in real time for improved clinical interpretation.
• Designed a FastAPI inference backend and Streamlit-based visualization dashboards to deliver GPU-accelerated predictions and interactive diagnostic insights.
• Improved model explainability by integrating attention-based visual cues, enhancing transparency and practitioner trust in automated diagnosis.

Technologies Used:

[PyTorch, FastAPI, Streamlit, Medical Imaging (DICOM/NIfTI), NumPy, TorchIO, PyVista]

Position:

Deep Learning / Medical AI Developer

Learnings and Benefits:

Advanced expertise in medical imaging, deep learning segmentation, 3D volumetric visualization, attention mechanisms, and GPU-accelerated inference for clinical diagnostic systems.

Description:

Trade AI is an AI-powered project designed to analyze trading charts, provide insights, and improve user experience in trading.

Technologies Used:

[Python, Data Analytics, Machine Learning, Trading API]

Position:

AI Developer

Learnings and Benefits:

Financial Market Analysis, Enhanced User Interaction...