full-stack web application that automates daily stand-up summaries by extracting relevant content from slack and jira logs using gpt-4 and langchain, processing 100+ messages per summary.

backend agents in python and fastapi that ingested and analyzed disaster data from 3 sources (social media, field reports, satellite imagery), processing 500+ data points per incident.

cloud-native system to verify digital image authenticity using 3 verification methods (perceptual hashing, invisible watermarking, cryptographic ledger), processing images up to 10mb with 93% accuracy.

end-to-end training pipeline that used gpt-4 to generate over 1,200 high-quality synthetic customer support tickets from 20 real examples, fine-tuned phi-3-mini to reach 96% accuracy.