For the longest time, I thought my path was in computer science. I was obsessed with coding since middle school, moving from HTML to competitive programming in C during high school. When I gained admission to the University of Dhaka, a teacher gave me advice that confused me: "You're good at math and coding; choose Statistics. We don't have enough people with that combination." I thought Statistics was just tally marks and pie charts. Even during my first year, I spent more time with robotics students solving C++ problems alongside fully engaging with my major.
The turning point came after my first year. While teaching myself R, I discovered Data Science. Everything clicked. My coding background wasn't a distraction—it was a superpower for analyzing data. That moment transformed my curiosity into a clear goal: combining computation with statistical rigor to solve real-world problems.
From Programmer to Data Scientist
My education at the University of Dhaka reflects my evolution from programmer to data scientist. I initially connected more naturally with courses that involved practical application than purely abstract theory. However, I excelled whenever computation was involved. I consistently earned top grades in my computational coursework and became the person peers sought out for implementation help. I frequently helped them translate statistical formulas into functioning code, which solidified my own understanding. By my final year, I had overcome my struggles with theoretical courses. This breakthrough was evident in my senior year performance where I achieved a 3.24 GPA, my A+ undergraduate thesis, and through my publications. This demonstrated my ability to integrate theory with computational practice.
Research Trajectory
My transition from coursework to research began with my undergraduate thesis on the Double Burden of Malnutrition in Bangladesh. I analyzed 4,866 child-mother pairs to examine the link between malnutrition and childhood immunization—a gap in existing epidemiological literature. The work introduced me to rigorous data cleaning and variable selection, and I learned to use statistical inference to understand counter-intuitive public health trends.
Eager to move into high-dimensional computational statistics, I co-authored three studies that required extensive simulation and algorithmic design. In one study, I addressed cellwise contamination in big data, designing simulation studies that demonstrated Adaptive Lasso based on Gaussian Rank correlation significantly outperforms traditional estimators. In another, I explored financial time-series analysis, implementing cointegration tests to model long-term equilibrium relationships. I also applied Dynamic Time Warping with hierarchical clustering to capture non-linear dynamics in price data that standard Euclidean metrics failed to detect.
My research trajectory led to three first-author papers (Q1-accepted) focusing on Hybrid Machine Learning Architectures for risk and security. In one paper, I designed a four-layered hierarchical framework to distinguish between fraud and operational errors. The system funnels data through an unsupervised Autoencoder, specialized XGBoost classifiers, and a Logistic Regression meta-learner. I tackled the precision-recall trade-off in mobile banking fraud detection by engineering a stacked ensemble model combining Bidirectional LSTMs for behavioral biometrics with gradient boosting for transactional features. I also examined standard Customer Lifetime Value models using Stacked Generalization and organically demonstrated that Loss Given Default is stochastic rather than deterministic. Additionally, I designed an analytical pipeline for cesarean section prediction using a Stacked Hybrid Model enhanced by SHAP interpretability.
Building StatClinix
Beyond publications, I sought to apply my skills in real-world settings. I co-founded StatClinix Ltd as CTO, where I led the data analysis and research for diverse clients, including government employees and medical professionals. This experience spans diverse fields, including FDIA detection in Smart Grids and MRI analysis for glioma grading, which has strengthened my research and data analysis capabilities across diverse problem domains.
Looking Ahead
My research interest lies in developing new statistical methods. I want to create frameworks that handle high-dimensional, messy data in resource-constrained environments. This interest stems from challenges I see in Bangladesh. In healthcare, we have one doctor per 1,300 people. I want to build AI systems that enhance physician efficiency and improve patient outcomes. I am also motivated by disaster preparedness to develop models that optimize logistics planning before and after natural disasters. I am equally drawn to precision agriculture to create systems that process real-time sensor data to make autonomous farming decisions, minimizing human effort while maximizing production and reducing climate-related losses.
After graduation, I aim to work as an academic or industrial researcher. I want to translate statistical theory into real-world solutions.