Hi, I'm Sakib
Data Science Graduate Student @ TU Dortmund
Data science enthusiast with Computer Science and Engineering background, focused on NLP and AI.
Contact MeAbout Me
My Introduction
Data science Graduate student with a Computer Science background, passionate about uncovering insights through AI and machine learning. Recently completed a master’s seminar at TU Dortmund on leveraging transformer models for depression detection and behavioral analysis in social media (Grade: 2.0). My bachelor thesis focused on real-time speech translation from Bengali to English, achieving a BLEU score of 48.98 (Grade: 1.3). As a former Research Assistant at the Economic Research Group (ERG), I worked on data analysis and developed predictive models. I found my interest in data, NLP, AI, and text analytics— now diving deeper into statistics and data analysis to broaden my expertise.
Completed
Written
Projects
Skills
My Technical LevelDevelopment
All About the CorePython
90%SQL
85%MS Excel
70%JavaScript
60%R
50%Android
40%Frameworks
Everyone Needs SupportNumPy
90%pandas
90%scikit-learn
85%Pytorch
85%streamlit
80%matplotlib
70%seaborn
70%OpenCV
65%NLTK
60%Flask
50%Machine Learning
Theory, theory!Ensemble Models
90%Exploratory Data Analysis
90%Natural Language Processing
85%Linear and Logistic Regression
80%Decision Trees
80%Convolutional Neural Networks
80%Recommender Systems
75%Clustering
65%Graph Neural Networks
60%Multi-modal Learning
60%Time Series
20%Cloud and Engineering
Fly Fast & High!Docker
60%AWS EC2
50%AWS ECR
50%Azure
40%Databases and Viz
Wow! FactorMySQL
85%Tableau
50%Power BI
50%Qualification
My Personal JourneyMasters of Science in Data Science
Technical University of Dortmund, Dortmund, GermanyExchange - Data Science (ongoing)
Ruhr University of Bochum, Bochum, GermanyB.Sc. in Computer Science & Engineering
North South University, Dhaka, BangladeshHigher Secondary in Science
Chattogram, BangladeshResearch Assistant
Economic Research Group
-
Conducted comprehensive data analysis using Python to support economic research projects.
-
Utilized MLflow to develop end-to-end predictive modeling suite, improving forecasting accuracy by 15%.
-
Led the implementation of ML workflows, reducing policy recommendation processing time by 2 days.
Research Assistant
North South University
-
Co-developed real-time speech translation system achieving BLEU score of 48.98 using Transformer architecture.
-
Executed data preprocessing, model training, and hyperparameter optimization for NLP system.
Portfolio
My Projects
Bayesian Approach For Smoking Prediction
Bayesian Data Analysis
Address the prediction of smoking behavior using individual-level factors (age, gender, income) and regional variation.
Age was negatively correlated with smoking — young individuals were more likely to smoke.
Income was nonlinear in trend — both the top and bottom income groups had high smoking prevalence. Education mattered — but its influence varied significantly by regions.
Tech Stack
Research Papers Referred
Bürkner, P.-C. (2017). brms: An R package for Bayesian multilevel models using Stan.
Loo: Efficient leave-one-out cross-validation and waic for bayesian models.
R interface to Stan
Transformer Decoders with MultiModal Regularization for Cross-Modal Food Retrieval
R: A language and environment for statistical computing
Pneumonia Detection Classification with MLFLOW DVC and CI/CD
🧠 MLOps (Machine Learning Operations)
Utilized Continuous Integration/Continuous Deployment (CI/CD) workflows to automate model tracking, versioning, and deployment on AWS using EC2 and ECR.
Tech Stack
View Code
Kidney Diseases Classification MLFLOW DVC
Computer Vision
Utilized Continuous Integration/Continuous Deployment (CI/CD) workflows to automate model tracking, versioning, and deployment on AWS using EC2 and ECR.
Cut down manual intervention by 40% with the integration of Docker and self-hosted runners for continuous deployment. Enhanced experiment tracking efficiency by 25% using MLFLOW for logging and managing model metadata.
Tech Stack
View Code
End To End Resume ATS Tracking LLM Project With Google Gemini Pro
Natural Language Processing (NLP)
Developed and implemented an end-to-end ATS resume scanning system using LLMs, incorporating Google Gemini Pro. Optimized tracking and filtering for enhanced candidate screening, improving accuracy and efficiency in the hiring process.
Deployed online via Streamlit
Tech Stack
HR Analytics
Data Science
Developed a dashboard in Power BI to analyze three months attendance data of AtliQ Company for HR manager.
With an accuracy of approximately 94.04% in April, AtliQ outperforms the other three months.
Approximately 93.21% of Mondays are the most active for AtliQ to enable them to hold their Monday meetings.
Tech Stack
Messenger Chat Analyser
Exploratory Data Analysis & NLP
The foundation of this app is made up of very basic data analysis, like adding up the times of day that messages were sent or counting the number of messages sent each month.
Implemented Sentiment Analysis
Tech Stack
Speech Recognition with python and Goolgle cloud api
Natural Language Processing
Speech Recognition From Bengali audio to text.
It's a part of Senior Design project in North South University.
Improved the performance of the application by implementing the transformer architecture as a machine translation (BLEU 48.98) model that gives a real time translation.
Tech Stack
Research Papers Referred
View Code
COVID-19 Spread Forcasting in Bangladesh on weekly and monthly basis.
Data Science (Machine Learning - Time Series Forecasting)
Prophet is a procedure for forecasting time series data based on an additive model where non-linear trends are fit with yearly, weekly, and daily seasonality, plus holiday effects.
Prophet paper: Sean J. Taylor, Benjamin Letham (2018) Forecasting at scale. The American Statistician 72(1):37-45 this paper
Tech Stack
Portfolio App
Android Development
Developed an Android application that acts as a portfolio of myself
It's a fun project to develop my skills.
Tech Stack
Certifications
Extra Courses I have Undertaken
C# Fundamentals for absolute beginners
Expiry Date: Does not expire
Create Your First Web App with Python and Flask
Expiry Date: Does not expire
Natural Language Processing with Attention Models
Expiry Date: Does not expire
Improving Deep Neural Networks: Hyperparameter Tuning, Regularization and Optimization
Expiry Date: Does not expire
Neural Networks and Deep Learning
Expiry Date: Does not expire
Structuring Machine Learning Projects
Expiry Date: Does not expire
Artificial Intelligence
Expiry Date: Does not expire
Blog
My Technical Articles
Can a Bayesian Model Predict Who Smokes?
Using Probabilistic Thinking to Decode Health Behaviors.
Smoking remains a significant global public health problem, and more insight into the demographic and socioeconomic determinants that influence smoking status is required to design effective, targeted interventions.
Read it!
Can AI Decode Silent Cries for Help? Transformers vs. Machine Learning in the Race to Detect Depression
🌐”I don’t want to live anymore.”
🌐 ”I’ve been feeling empty for weeks.”
🌐 ”Why does everything hurt?”
These aren’t just social media posts — they’re digital SOS signals from 280 million people battling depression worldwide.
As suicide claims 700,000 lives annually, researchers are deploying AI as a linguistic lifeguard.
But which approach dives deeper: transformer models or traditional machine learning?
Let’s dissect two groundbreaking studies reshaping mental health tech.
