Research
Flood Event Extraction from News Media to Support Satellite-Based Flood Index Insurance in Bangladesh
Natural Language Processing Researcher
Researched on "Flood Event Extraction from News Media to Support Satellite-Based Flood Index Insurance in Bangladesh". Created and published a dataset of 40,000 tagged news articles covering flood events in Bangladesh by 10 prominent news sources. Developed a BERT-based classifier to extract flood-events. Created a flood-event time-series and defined criteria for flood occurrence and severity. Validated results against Sentinel data, with a correlation coefficient of 0.7. The results contributed to the development of a flood index insurance by the Bangladesh government. Authored a pre-print research paper and presented findings at the AGU conference.
DVMM Lab
Computer Vision Researcher
Developed a phrase grounding pipeline using YOLOv3 and BERT for image and caption extraction from research papers. Achieved 85% accuracy. Constructed a searchable knowledge graph between image and captions. Classified Dosage Response curve using extracted features. Achieved 92.7% accuracy with AdaBoost.
SMARTtest: A Smartphone App to Facilitate HIV and Syphilis Self- and Partner-Testing, Interpretation of Results, and Linkage to Care
Full Stack Developer
Created an affordable HIV & Syphilis detection app using React Native and Firebase, with Twilio & SendGrid for data sharing. Automated testing & deployment at Expo. The app has been downloaded 1000+, with news coverage. Published research in AIDS and Behaviour.
INCITE Labs
Data Science Researcher
Extracted syllabi and mission statements from college websites to quantify measure for liberal arts education. Developed Python scripts for streamlined SQL database interaction.
Pill Recognition & Prescription Extraction
Machine Learning Researcher
Utilized Google Vision and OCR to extract pill features and bottle imprints. Developed a multi-dimensional embedding using collected data for RandomForest and SVM classifiers, enabling precise pill identification.
A Study on the Solar Illumination Provided by a Water Bottle
Researcher and First Author
Experimentally demonstrated that the “Liter of Light” bottle outperforms a glass plate in illuminating low-light areas like slums. Regional Finalist for Google Science Fair. Published and presented research at Journal of Basic and Applied Engineering Research. Received media coverage for my research, with features in The Tribune, The Times of Indiaand Krishi Jagran national newspapers.