About

Smaranjit Ghose is a Technical Product Manager at Twinkle Twilight with a passion for building data-driven solutions for healthcare, finance, and social problems. He has a strong background in statistical analysis of data, identification of patterns and biases, and implementation of predictive models algorithmically. Currently pursuing a Masters in Data Science program at Columbia University, Smaranjit's research interests lie in Computer Vision, Explainable AI, and Biomedical Imaging. Smaranjit is a lifelong learner and has volunteered as a project admin at 17 open-source initiatives to mentor students in their initial open-source journey. He has impacted 6000 mentees in the process and has authored 9 research papers on the application of Deep Learning to biomedical imaging. He has successfully presented them at international conferences. Smaranjit has previously worked as an intern at UCL CANDI, IIT Madras, and SRMIST. He has volunteered as a hackathon judge and mentor for 48 international hackathons to provide valuable feedback on projects related to sustainable development, women empowerment, and digital healthcare. Currently, Smaranjit is a Fellow at ACM Association for Computing Machinery Winter School 2020-21. He has studied the impact of Fairness, Accountability, and Transparency in Artificial Intelligence in the Winter School hosted by IIT Kharagpur and sponsored by IBM Research India. He has participated in group projects to determine the possibilities of integration of privacy-preserving methods into small scale projects using PySyt. In his past experience, Smaranjit has worked as a Research Intern (Deep Learning) at UCL, where he developed an image classification pipeline using TensorFlow to identify the make and model of knee implants resulting in an achievement of 96.72% validation accuracy. He has also coordinated the collection of 50000 radiographs from hospitals in India and the United Kingdom by creating and maintaining a non-relational database utilizing Firebase, thereby diminishing staff time for the process by 65%. Smaranjit has also worked as a Data Science Intern at Tripura University (a central university), where he devised image segmentation models utilizing U-Net with PyTorch for separation of probable carcinogenic regions from the image for screening of cervical cancer with an intersection over union score of 0.9. He has also deployed a chatbot leveraging RASA and DialogFlow for assisting 180 patients in the collaborating medical facilities with an easily accessible platform to clarify diagnosis or therapy-related queries