Paras Adhikari

I built an end-to-end voice cloning pipeline using Python, PyDub, and Hugging Face Transformers, preparing timestamped audio-to-text datasets and integrating transformer-based models for voice synthesis.

For my multilingual named entity recognition project, I moved from a spaCy baseline to a multilingual BERT model for English, German, and Turkish data. I reduced noisy output with confidence thresholding, regex cleaning, and false-positive filtering.

My projects also include cleaning and analyzing crime data, building a Binance Futures trading bot, and creating a house-price prediction model. I worked as a Python Developer Intern at Skill India and supported retail operations through data handling and reporting as a Cluster Planner (CP) – MIS Executive Junior at V2 Retail Limited.