Applied Scientist, HERE technologies (Aug'23-Current)
Working on the research and development of computer vision and language-vision multimodal models for extracting real world navigation specifications and attributes from varied sensor data sources for map construction.
Doctoral Researcher, IITH (Sep'20-Nov'23)
Research focus at the intersection of Representation learning for static and dynamic graphs, spectral graph theory and topology along with applications to Knowledge Graphs and recommendation systems.
Data Scientist, Servify (Oct'18-Jul'23)
Designed and implemented a CNN and Transformer-based system, for extracting multi-modal information (objects and relevant text) from 1Mn+ images uploaded by users.
Implemented semantic segmentation models for detecting various kinds of damages on phones. Our model uses both top-down and bottom-up instance segmentation approaches, and learns maps for each instance using convolution layers.
Designed and developed a model for large-scale classification and entity extraction from documents. Model is implemented to handle four languages including English, German, French, and Dutch.
Implemented a data processing pipeline to automate training and validation of models to ensure continual learning from user-generated temporal data.
Worked on the research and development of a graph neural network-based system to predict product vulnerabilities and find correlation with similar products to help decision making for new product plans.
Built models to convert unstructured documents into structured format (Knowledge Graph) for question answering capabilities.
Researched and built prototype to provide empirical evidences to switch product technology from traditional ML techniques to the latest large language-models.
Built damage detection models from limited data to automate detection of aberrations from images of devices obtained from a physical kiosk.
ML Engineer, TCS (Jun'16-Oct'18)
Developed a classification system for automated mail reply using tensorflow on IBM lotus notes
Developed a conversational agent for e-commerce product recommendations using sequential information
Designed and developed a conversational agent for answering generic FAQs
Worked on the research and development of a geo-positioning and tracking system that automatically determines and bills a user's journey using ML, for a client in the Netherlands.
For further details, please find my resume attached here