Trisha Sengupta

I am a machine learning engineer with hands-on experience building and deploying models for audio classification and time-series forecasting. I engineered a 10-genre music classifier using CNNs and Transformer architecture that achieved 93% accuracy and developed SARIMAX time-series models capturing monthly seasonality with strong fit metrics.

I focus on feature engineering, end-to-end pipelines, and model deployment, improving accuracy by 8% through optimal audio feature selection and designing data-driven pricing strategies using demand elasticity. I contribute to hackathons and product-focused projects—building an autonomous fire-detection robot and a BERT-based customer query routing pipeline with real-time dashboards—thriving in collaborative teams that translate insights into production-ready solutions.