Tawsif Mahmud

Building the Full Stack Powered by AI.

From crafting user-friendly web applications and resilient backend services to implementing sophisticated Machine Learning algorithms, I deliver complete, high-performance software solutions. My work bridges the gap between user needs and advanced AI capabilities.

Tawsif Mahmud - Profile visual representation

Technical Proficiency

A showcase of the tools and technologies I work with.

Programming Languages

Python
JavaScript
TypeScript
C#
Java
Dart
HTML5
CSS3

Frameworks & Libraries

React
FastAPI
Angular
Flutter
Django
ASP.NET
Tailwind CSS
Bootstrap
Streamlit

Machine Learning

TensorFlow
PyTorch
OpenCV
LangChain
LLMs (Llama 2)
Scikit-Learn
Pandas & Numpy
Airflow
Ollama

DevOps, Cloud & Miscellaneous

Docker
Git & GitHub
Azure
MongoDB
PostgreSQL
Superset
Keycloak

My Portfolio

Explore a selection of my projects, filterable by category to see specific areas of my expertise.

View All Projects List
Vision Inspection System

Vision Inspection System

Manufacturing Automation

TensorFlowReactFastAPI

Real-time, automated quality inspection using computer vision and deep learning for caravans.

Client: Australia

Local LLM Study Assistant

Local LLM Study Assistant

Local LLM Information Retrieval

LangchainStreamlit

Robust RAG system for context-aware answers from documents using local LLMs.

GitHub
Edge Measurement

Edge Measurement

Image Processing to measure edges

Y

Real-time Edge Measurement using YOLOv8 and OpenCV for accurate measurements.

ML Training Pipeline with Airflow

ML Training Pipeline with Airflow

Automated Model Training

PostgreSQL

Automated ML pipeline with Apache Airflow for Iris dataset classification.

GitHub
AMD Detection using CNN

AMD Detection using CNN

Age-related Macular Degeneration Detection using existing and custom CNN models

Custom CNN architecture for Macular Degeneration detection with 95% accuracy.

GitHub
Caravan Ball Weight Estimator

Caravan Ball Weight Estimator

ML based Ball Weight Estimator

P
N

Based on caravan models furniture placement deatils, our ML model estimates the ball weight for safer manufacturing.

Client: Australia

My Journey

Professional
Academic
Jun 2024 - Ongoing

Software Engineer (ML)

Bangladesh Software Solution

Jan 2023 - May 2024

Research Assistant

North South University - Center for Applied & Computational Sciences (CACS)

Jan 2022 - Nov 2023

Undergraduate Teaching Assistant

North South University - Department of Mathematics and Physics

Feb 2021 - Mar 2022

Jr. Software Engineer

Accelerating Intelligence

Jan 2019 - Jun 2023

B.Sc. in Computer Science and Engineering

North South University, Bangladesh

Highlights:

  • CGPA: 3.80/4.0 (Summa Cum Laude)

Research & Publications

Thesis

My undergraduate thesis research in machine learning and computer vision.

Cross-Attention Diffusion Model for Conditional Image Generation

Undergraduate Thesis

Specializing in diffusion models, latent spaces, autoencoders, and variational autoencoders. Explored image generation, incorporating cross-attention mechanisms within the model, utilizing MNIST and Fashion MNIST datasets.

Technologies & Keywords:

Diffusion ModelsCross-AttentionLatent SpacesAutoencodersVAEMNISTFashion MNIST
Cross-Attention Diffusion Model for Conditional Image Generation - Demo

Model demonstration

Publications

A selection of my peer-reviewed articles and conference papers.

Artificial Neural Network (ANN) approach in predicting the thermo-solutal transport rate from multiple heated chips within an enclosure filled with hybrid nanocoolant

Mahmud, T., Saboj, J. H., Nag, P., Saha, G., Saha, B. K.

International Journal of Thermofluids, 2024, 100923

Keywords:

Artificial Neural NetworkThermo-solutal transportHybrid nanocoolantHeat transfer
View Publication
Artificial Neural Network (ANN) approach in predicting the thermo-solutal transport rate from multiple heated chips within an enclosure filled with hybrid nanocoolant - Figure 1
Artificial Neural Network (ANN) approach in predicting the thermo-solutal transport rate from multiple heated chips within an enclosure filled with hybrid nanocoolant - Figure 2

Convective heat transfer efficacy of an experimentally observed non-Newtonian MWCNT-Fe3O4-EG hybrid nanofluid in a driven enclosure with a heated cylinder

Mahmud, T., Nag, P., Molla, Md. M., Saha, S. C.

International Journal of Thermal Sciences, 2024, 204

Keywords:

Hybrid nanofluidConvective heat transferNon-NewtonianMWCNTFe3O4
View Publication
Convective heat transfer efficacy of an experimentally observed non-Newtonian MWCNT-Fe3O4-EG hybrid nanofluid in a driven enclosure with a heated cylinder - Figure 1
Convective heat transfer efficacy of an experimentally observed non-Newtonian MWCNT-Fe3O4-EG hybrid nanofluid in a driven enclosure with a heated cylinder - Figure 2

Entropy production associated with magnetohydrodynamics (MHD) thermo-solutal natural convection of non-Newtonian MWCNT-SiO2-EG hybrid nano-coolant

Mahmud, T., Chowdhury, T., Nag, P., Md. Mamun Molla

Heliyon, 2024, 10(15)

Keywords:

MHDEntropy productionHybrid nano-coolantNatural convectionMWCNT
View Publication
Entropy production associated with magnetohydrodynamics (MHD) thermo-solutal natural convection of non-Newtonian MWCNT-SiO2-EG hybrid nano-coolant - Figure 1
Entropy production associated with magnetohydrodynamics (MHD) thermo-solutal natural convection of non-Newtonian MWCNT-SiO2-EG hybrid nano-coolant - Figure 2