Abhishek Bhat

I’m a Python developer with hands-on experience building real-world automation, including statutory payroll, attendance, and deduction register workflows. I used Pandas and OpenPyXL to generate advanced reporting, formatting, and compliance checks.

I also build machine learning systems end to end, from multimodal feature extraction to model training and evaluation. One highlight is a multimodal price prediction approach combining Sentence-BERT and EfficientNet-B0 embeddings, trained with LightGBM regression using 5-fold cross-validation and early stopping.

Beyond tabular and text, I enjoy working with vision and 3D data pipelines. I’ve built an end-to-end 3D mesh workflow using Trimesh and Open3D (loading, normalizing, quantizing, reconstructing) and validated quality using reconstruction error metrics; I’ve also developed a hand-gesture gaming system using real-time computer vision.

My research focus is practical NLP, especially semantic parsing through adaptive data augmentation. I fine-tuned T5-base across augmentation strategies on the Parallel Meaning Bank, achieving strong BLEU and Token F1 results and contributing work accepted at IEEE CE2CT 2026—while staying driven by continuous improvement through competitions and collaboration.