Aravind Medapatla

I've built end-to-end AI systems for speech and emotion analysis, including a multimodal emotion recognition pipeline in PyTorch that achieved 100% test accuracy (F1=1.00) across seven emotion classes on 1,120 held-out samples.

I developed an Accent Intelligence System using HuBERT, MFCC feature extraction, and Gradio for real-time speech-based accent predictions. I also fine-tuned BERT for contextual text modelling and produced evaluation artefacts including confusion matrices, t-SNE plots, per-class F1 tables, and training reports.

Alongside AI work, I built Memory Explorer, a full-stack travel journal using Spring Boot, MongoDB, RESTful APIs, and role-based access control. I'm seeking an entry-level role or internship where I can build scalable applications and grow in real-world technology environments.