I built a speech emotion detection system in PyTorch that classifies six emotions from speech, using speaker-independent evaluation and achieving a 0.80 Macro AUC. I also integrated speech transcription and low-latency FastAPI inference.
For Customer Support AI, I developed an intent detection and grounded RAG platform that achieved 88.72% test accuracy across 77 intent classes. It retrieves answers from policy documents and includes security guardrails.
I also built an explainable medical computer vision pipeline using the RSNA dataset, fine-tuning EfficientNet-B0 and integrating Grad-CAM. My projects include movie recommendations and retail sales forecasting; as a Data Analytics Intern at UptoSkill, I cleaned and analyzed datasets and designed Power BI and Excel dashboards.

