Mandip Thakur

I’m an AI Fellow at Fusemachines, driven by building end-to-end data pipelines that move smoothly from ingestion to transformation to serving. I focus on modular, structured pipeline design so the same data quality that powers models also supports reliable downstream use.

During my fellowship, I worked with structured and unstructured datasets to support transformation and feature engineering aligned with downstream modeling needs. I also deployed data-driven applications using REST APIs and Streamlit, while applying core MLOps ideas like experiment tracking, data drift detection, and model monitoring for better pipeline observability.

Before that, I served as an Undergraduate Research Assistant with the RDI, Dept. of CSE at Kathmandu University. I built an OCR pipeline for Nepali text extraction from images and PDFs, using NLP preprocessing and data cleansing to improve data validation, extraction accuracy, and overall data quality—supported by field visits for primary data acquisition.

Alongside my work, I’ve delivered Azure end-to-end data engineering projects using ADF, ADLS Gen2, Databricks, and Synapse, including Bronze/Silver/Gold SQL patterns and solutions with security and monitoring components. I also explore forecasting and analytics through time series (ARIMA/SARIMA), customer segmentation (RFM), and simulation-based modeling with Python, supported by certifications such as DeepLearning.AI courses for medical diagnosis and prognosis.