cv

Dissertation: Human-Aligned Video-AI for Public Safety. Advisors: Prof. Cees Snoek, Prof. Marie Lindegaard.

Thesis: Comparative Analysis of ML Models on Real-World Building Datasets (Dr. Victor de Boer), 8.5/10. Two-year programme completed in 10 months · Top 5% of graduates.

Thesis: AI and Remote Sensing: Transferring Learned Patterns to Map Agriculture at Scale (Dr. Debayan Gupta, Dr. Meghna Agarwala). GPA 3.86/4.00 · Magna Cum Laude · Ashoka Scholars Programme. First student in the university's history to conduct research in the first year, with an exemption to take a third-year research course as a first year.

Video AI for public safety, with a focus on aligning video models to human judgement. Supervised by Prof. Cees Snoek and Prof. Marie Lindegaard.

Neuro-symbolic learning combining graph neural networks with knowledge graphs — building KGs from a learner's existing knowledge and using analogy to link it to target concepts.

Combined multi-temporal, multi-satellite optical, radar and topography data to map tea plantations across India — including smallholder mixed-cropping farms — and identify where they intersect elephant migration paths, supporting WWF work to reduce dangerous human–elephant encounters.

Developed an ML method combining multi-spectral, multi-temporal, multi-satellite (radar and optical) data with Random Forests and SVMs to map every mango farm across Uttar Pradesh from 30 m open-source imagery, and designed strategies with farmers to coordinate regional crop growth and prevent price crashes. Awarded Best Undergraduate Thesis.

Analysed ant self-organisation across competing paths to a food source to derive principles for optimising traffic flow.

Built and deployed ML pipelines to optimise gas and energy usage and automate sensor fault detection across hundreds of buildings in the Netherlands, productionising the thesis research below.

Established whether ML models generalise across buildings to forecast indoor climate — training LSTMs, CNNs, MLPs, Random Forests and KNNs on >9M datapoints and >800 features from 8 commercial buildings — and derived the sensor and data benchmarks required for reliable predictive control.

Led the e-Governance intern team in building a dashboard to collect and visualise data entered by officials, to detect and prevent financial fraud.

Built databases and ran data analytics to measure and extend the impact of a UNDP/IUCN-recognised venture supplying low-water hydroponic fodder systems to farmers in harsh climates.

Proof-of-concept LSTM model predicting food surplus to optimise redirection to nearby shelters, orphanages, food banks, animal feed, composting or combustion. Awarded Best Entrepreneurial Idea by VU faculty out of 33 ideas; admitted to the Demonstrator Lab venture incubator.

Founded and organised the university's first WiDS conference and ran it for two consecutive years; the second year's livestream drew over 1,000 views. Interviewed leading women in data science including Gargi Dasgupta (Director & CTO, IBM South Asia), Ujjyaini Mitra (CDO, ZEE Global) and Radhika Kulkarni (President, INFORMS).

Developed the ACM IKDD website and initiated events and resources to make AI and data science more accessible.