Akshat Pradhan

Developing facial & emotional recognition model for the company, through training a YOLOv8 model.

A network of high-achieving high school students with the purpose of forming new connections and building off of each other. Our members are founders and creatives with the brightest futures of any students.

Working on a research project regarding computer vision and facial emotional recognition in the Inspirit AI Scholars Program - 1v1 mentorship.

An AI-powered healthcare tool designed to help hospitals and clinics manage patient flow more effectively. It combines real-time data (like admissions, discharges, and bed usage) with analytics and AI insights to improve resource allocation, reduce wait times, and support better decision-making for both patients and providers. (Still in Development)

Developed an AI-powered application designed to help neurodivergent individuals, including those with ADHD, autism, etc, navigate social situations with confidence. The app takes input in text, audio, and video, analyzes it using natural language processing and computer vision, and provides real-time suggestions for socially appropriate responses. (Video component is still under development)

Joined the Erdos research group under Oxford professor Dr. Ajit Jaokar, focusing on the intersection of AI, machine learning, and efficiently developing AI applications.

Learned the fundamentals of computer vision model development, including dataset preparation, model selection, and training workflows. Gained hands-on experience in training deep learning models for facial expression and emotional recognition. Worked with labeled image datasets, ensuring data quality and variety for better model accuracy. Implemented preprocessing techniques such as resizing, normalization, and augmentation to improve model performance. Explored different neural network architectures suitable for emotion detection. Understood evaluation metrics (accuracy, precision, recall, F1-score) to assess model performance. Learned model optimization techniques like fine-tuning, hyperparameter tuning, and reducing overfitting. Studied real-time inference possibilities to integrate facial/emotional recognition into interactive applications.

Developed applications that incorporate and call external APIs, using tools like Postman and RESTful APIs, making sure that there is smooth communication between the front-end and back-end services, with a focus on raising the performance and functionality of Flutter-based apps. - Frontend Development

Taught neurodivergent children (8-10yrs old) Math and English.