NLP Engineer | Deep Learning
Academic Portfolio · Oct 2025 – Present
- Specialized in transformer-based NLP applications with focus on question answering and information extraction systems using state-of-the-art architectures.
- Developed production-ready extractive QA system using fine-tuned BERT on SQuAD dataset.
- Key Achievements:
- • Fine-tuned BERT-base (109M params) achieving 95% answer extraction accuracy
- • Implemented robust tokenization pipeline with 100% answer position validation
- • Trained model in 4.6 minutes on Tesla T4 GPU with early stopping (validation loss 1.65)
- • Built dual web interfaces (Gradio with public sharing + Streamlit dashboard)
- • Created 83 pages of technical documentation across 4 comprehensive notebooks
- • Achieved 50ms inference time enabling real-time applications
- Technologies: BERT, PyTorch, Hugging Face Transformers, SQuAD, Gradio, Streamlit, NLP, Transfer Learning







