About Me

Hello, I'm M S Varun. I have a passion for app development, web development, and machine learning.

Experience

Research Intern • Interned in a research lab at PES University - CDSAML (Center for Data Science and Applied Machine Learning) • Our team researched the topics – Meta Learning in Deep Learning and built a project about multimodality and classification using less pair-wise labelled data. • Drafted a research paper in IEEE format and will be published soon.

Projects

Meta Learning to assist Classification with Minimal Data

• Developed an ML model that can perform classification using image and text data at a decent accuracy with limited number of pair-wise labelled data. • Meta Learning was used to demonstrate how the model optimizes its meta parameters with less available labelled data (in the case of social media data). This was simulated in our research. • We tested two approaches of meta learning – MAML (Model Agnostic Meta Learning) and CAVIA (Fast Context Adaptation via Meta Learning). The latter proved to perform better in our case.

Multimodal Six-Way Classification via Fusion of CLIP Embeddings

• Developed an ML model which performs bimodal classification (image and text) among six classes. The model was deployed on a website built using Streamlit. • The image and text data were converted to CLIP (Contrastive Language Image Pretraining) Embeddings. • The embeddings were formed using the respective encoders (text and image encoders). The fusion/concatenation of these embeddings were used for training the model. • Use of CLIP embeddings enabled us to achieve 12% more accuracy compared to the conventional image and text preprocessing techniques (like OpenCV and NLP).

Unified Medical Interface (UMI)

• Developed a mobile application (prototypical) using Flutter and Firebase, to book an ambulance from the nearest hospital. • Made use of Machine Learning algorithms to dynamically locate the nearest hospital from the user’s current location. • The app uses a map API from MapBox which is used to present the number of remaining beds in each hospital around the user’s locality. • The app will also suggest remedies in case of mild emergencies or injuries

Socials

LinkedIn: M S Varun

GitHub: varun80042