Md. Saddam Hossain

I started Kite Game Studio Ltd., an in-house software development company working for image processing app for iOS and Android platforms. Currently, I am working on Retouch app where I am utilizing Flux.1-dev a cutting-edge genAI model for cloth replacement, Hair-style generation, Lipstick style generation. I integrated Flux.1-dev model with Controlnet. Using Segformer a model for human face parser for mask generation.

I joined Braincraft Ltd. as a Senior MLE. BCL is an image processing in-house software development company where we build product for iOS and Android platform as well as web platform. I joined in AIML R&D team. My role was to build ML models from scratch and fine-tune the in some cases, and doing R&D about it for making develoyment pipeline. Model training, optimization, Fine-tunig, and api development as well. I ful-fledge training and fine-tuning u2net model from data preparation to data pipeline doing model surgery. Training the model using 20k, 30, 40 and 50k samples etc phases. I also applied different augmentation techniques for the improvise the datasets. I make surgery in tiny u2net model which originally was 4.7MB and I increase the size to 14.5MB for mobile platfrom. You can check and use the product from Playstore. I also work for cartoon.ai app where my role was to develop the face-swapper model deployment pipeline where I use GFP-GAN and REAL-ESRGAN. It is work for both image to image and image to video swaping also. Besides, I also doing R&D about cartoon style generation using Animegan3. Also work for stable diffusion webui for hair style and dreambooth model style generation. Deploy the webui pipeline in replicate platfrom.

I was joined as a machine learning engineer at BJIT Ltd.. At BJIT, I learned and applied classical ML as well as deep learning technique. I learned different machine learning library such as sk-learn, pytorch, tensorflow and some sort of visualization tools such as matplotlib, seaborn etc. I was trained for different regression model, classification model such logistic regression, decision tree, random forest, svm, and clustering method such k-means and kmeans++. Introduced also neural network ANN and DNN. I also learn and used CNN like alexnet, lenet, googlenet etc for cats and dogs datasets.

I was joined as a full-stack Java web developer at Orbund LLC where I leaned web concept, full-stack development using raw javascript and HTML, CSS. I also used mysql database.

BSc at the Shahjalal University of Science and Technology. This is where I first got into deep learning, attending ACM Labs's for competitive programming and reading groups.