At Hydroponic GreenTech Solutions, I delivered SweetCherry, a decision-support platform for cherry growers piloted with SPP-C by farmers and experts across five regions. It combined disease diagnosis, price forecasts, and weather signals.
For SweetCherry, I built LightGBM quantile and SARIMAX price forecasts with weather inputs, testing them on simulated prices because real market data was sparse. I also found a gap between a MobileNetV2 classifier's Kaggle results and its performance on farmers' phone photos, then retrained it with expert-labelled field photos.
At Islamic Development Bank, I reconciled transaction records across source systems using Python and SQL checks. The checks matched records and flagged mismatches, gaps, and duplicates for compliance review.
In my Strata project, I built a retention-targeting score for a retail bank and a tool to project branch workload under migration campaigns. With DetectionX5, I developed an offline plate-reading platform for vehicle intake, including confidence signals and a review queue for doubtful reads.
