cs + econ @ barnard columbia. i build things at the intersection of AI, genomics, and finance — and then i go do bhangra. currently researching quantum-informed imaging pipelines by day, figuring out the best ramen spot in manhattan by night.
sophomore at barnard columbia (class of '28) studying CS & economics. i came in pre-med, pivoted to CS, and never looked back. i genuinely love what i do — whether that's debugging a pytorch model at midnight or presenting investment decks to partners. i'm also sikh, so community and music (kirtan 🙏) are a big part of me outside the terminal.
AI + quantum research at GILMLab, columbia. building GPU-accelerated imaging pipelines using quantum optical models. it's as wild as it sounds.
new york city — morningside heights specifically, but you'll find me in chinatown, flushing, or basically anywhere with good food.
developing AI pipelines with quantum optical models for multispectral imaging. applying deep learning to non-invasive biological signal detection. translating quantum principles into scalable, GPU-accelerated algorithms.
built predictive drug sensitivity models (XGBoost vs. Random Forest) on 578 cancer cell lines and 50 drugs using CRISPR-Cas9 data. XGBoost won. published the results.
researched redevelopment of a 200K+ sq. ft. shopping center. built 3+ investment pitch decks for senior partners on million-dollar acquisition scenarios. financial modeling, ROI analysis — the works.
supported ML + cybersecurity projects alongside 5+ consultants. helped evaluate $500K+ of RFPs and contributed to 10+ technical review meetings.
7+ presentations to 100+ peers. won Best Paper Award (distracted driving project) out of 50+ submissions. still the thing i'm most proud of from high school.
XGBoost + Random Forest models predicting drug sensitivity from CRISPR gene knockouts across 578 cancer cell lines. identified top predictive genes for drug response. it's in a journal now which still feels surreal.
real-time ASL-to-text using TensorFlow Object Detection + transfer learning on a custom-labeled dataset. built and deployed as a working web app.
5 CNN models trained on 7,400 images to classify distracted vs. focused driving. best paper award out of 50+ submissions.
regression models forecasting turbine output from long-term data. full pipeline: imputation, visualization, and model evaluation (RMSE/MAE/MSE).
devotional music is a grounding force for me. being sikh shapes how i show up in community, in work, in everything.
high-energy punjabi folk dance. it's the best workout i've found and it connects me to culture in the most joyful way possible.
flushing for xiao long bao, chinatown for soup dumplings, everywhere else for whatever is new. i am always looking for recommendations.
started my GILMLab rotation and the stuff we're doing with quantum fluorescence models is genuinely blowing my mind. never thought i'd be writing GPU-accelerated algorithms that model light-matter interactions but here we are. also had the best khao soi in jackson heights this week — coincidence? i think not.
the genomics paper from children's national is officially out. the moment i saw my name on the frontiers site i had to sit down. this is what two semesters of debugging pytorch at 2am looks like. worth it.