- Design and develop a multilingual customer service chatbot using Python, Rasa framework, and multilingual transformers.- Curate, clean, and organize data in multiple languages to ensure chatbot accuracy and relevance.- Develop user-friendly tools for dataset curation and non-technical staff to manage and control the chatbot.- Address NLP challenges for low-resource language Tajiki, ensuring the chatbot has good Natural Language Understanding.

- Investigated transformer-based models for information retrieval- Effective pre-training and fine-tuning objectives and maximizing training and inference efficiency- Adapt transformers architectures to large-scale CPU-only applications- Conducted two literature reviews on multimodal document ranking and sparse attention mechanisms for short text sequences- Used Pytorch, PyTorch Lightning, Huggingface models, Pandas, Numpy, Sklearn, Docker.

- Studied the latest literature on speech synthesis- Trained transformer-based architectures for text-to-speech synthesis- Used Pytorch and Yandex's internal tools for executing and debugging long-running experiments

- Developed deep learning optimization pipeline using PyTorch Lightening- Utilized Git for version control and collaboration- Implemented data preprocessing and architecture selection for optimal performance- Integrated accuracy metrics and various loss functions for diverse tasks in classification, face recognition, and semantic segmentation

Trained high school students for national and international-level mathematics and programming competitions. Actively participated in conducting national team selection exams for international competitions.