Balyogi Mohan Dash, PhD

Experience

• 

  Indian Institute of Technology, Kharagpur

  Jul 2019 - Oct 2021

  Teaching Assistant
• 

  University of Lille 1 Sciences and Technology

  Jul 2021 - Mar 2024

  - Developed a deep learning method using Python and Tensorflow to estimate the presence of faults in critical industrial components 37% earlier than the traditional method. - Spearheaded the creation of a novel explainable artificial intelligence (XAI) method that enhances trust in predictions from black box models used in critical industrial settings. - Conducted proof of concept on renewable energy systems, and presented findings at an international conference. - Recognized the lack of available content on data-driven condition monitoring on YouTube, and leveraged expertise in Python, Scikit-learn, Pandas, TensorFlow, and MATLAB to create a community-focused channel in this area.- Established a thriving YouTube community with 50,000+ views and 600+ subscribers, emphasizing communication and knowledge sharing while demonstrating proficiency in content creation for data-driven condition monitoring.

  • PhD in Artificial Intelligence

    Oct 2021 - Mar 2024

  • Educator on Youtube (1500+ subscribers)

    Jul 2021 - Mar 2024
• 

  BUAWEI

  Mar 2024 - now

  Research And Development Engineer

  - Developed a state-of-the-art vision anomaly detection model for real-time quality inspection in collaboration with a large automotive client, improving speed by 85% without compromising performance. Collaborated with DevOps for deployment and coordinated with the product team for post-production monitoring. Tech stack: PyTorch, CI/CD, Git, Docker, Linux. - Utilized diffusion models to generate a more diverse dataset, enhancing representation of underrepresented classes and improving model accuracy by 10-20%. Tech stack: Hugging Face, PyTorch, Linux, Transformers.- Worked on various generative AI projects involving text-to-image models, image editing using generative AI, inpainting, guidance, and stable diffusion techniques. Tech stack: Hugging Face, PyTorch, Linux, Transformers.- Led an object tracking project for an international aerospace company, delivering a real-time contamination estimation system that was 10x faster than manual methods. This method is now in production. Tech stack: OpenCV, Ultranalytics, PyTorch, CI/CD, Docker, Linux. Show less