I’m a Master’s graduate in Cognitive Science with a strong passion for LLM, Deep Learning, and Computer Vision. In my Master’s thesis, I explored advanced deep learning methods for fluorescence microscopy, focusing on AI-based denoising and reconstruction at low signal-to-noise ratios.

Alongside my academic journey, I have worked as a Research Assistant, Student Assistant, and Student Tutor, developing solid programming, debugging, and analytical skills. I’m eager to apply my knowledge to impactful projects in Computer Vision, LLM, GenAI, and Data Science.

• Programming Languages: Python, R, C++
• Deep Learning & Computer Vision: TensorFlow, PyTorch, Keras, OpenCV, Numpy, Scikit-Image, MONAI
• Data Science & NLP: Pandas, Scikit-Learn, NLTK, Transformer Architectures, Generative AI
• LLMs & AI Agents: LangChain, LangGraph, CrewAI, Gradio
• Tools: Git, GitLab, Docker, Visual Studio, MLFlow, W&B, HPC(SLURM)
• Operating Systems: Linux, Windows

Demo [https://github.com/ArghaSarker/mitosam-vit]

• Fine-tuned Segment Anything Model (SAM) for mitochondria segmentation using LoRA for efficient domain adaptation.
• Ran experiments comparing prompting strategies and measured their impact on segmentation quality and robustness.
• Used xAI techniques to interpret predictions, analyze failure cases, and strengthen confidence in model behavior to improve trustworthiness for biomedical image applications.

Demo [https://github.com/ArghaSarker/Ask-a-Nerd]

• Built a PDF-based chatbot that allows users to upload documents and ask questions about their content.
• Implemented a Retrieval-Augmented Generation pipeline using LangChain, OpenAI embeddings, and vector search to retrieve relevant document chunks.
• Designed an interactive Panel interface with conversation history, document configuration, and a database tab showing retrieved sources for better transparency.

Demo [https://github.com/ArghaSarker/llm_project]

• Compared the base model vs full instruction fine-tuned model vs PEFT fine-tuned model.
• upcoming: RLHF and model distillation

Read my thesis here: [https://github.com/ArghaSarker/Mather-Thesis-]

Demo [https://github.com/ArghaSarker/RDL_denoising]

Demo [https://github.com/ArghaSarker/projection_upsampling_network]

• Developed a robust SIM reconstruction model to enhance image resolution using deep learning.
• Achieved improved reconstruction speed and quality compared to traditional Fourier-based methods.
• Application: Low SNR fluorescence microscopy data in bioimage analysis.

Demo [https://github.com/ArghaSarker/Data_augmentation_with_VAE]

• Built a VAE to generate synthetic microscopy data, supporting improved deep learning training.
• Helped address challenges with limited high-resolution datasets in bioimage analysis.

Demo - Not available due to company ownership

project report - https://github.com/ArghaSarker/Protect_your_privacy

• Collaborated with LMIS GmbH to develop a privacy-preserving solution based on GDPR.
• Focused on anonymization of license plates, faces, tattoos, texts, and screens using DL models.

Demo [https://github.com/madammann/SaccadicEyeMovementNET]

• Engineered a pipeline to fetch, preprocess, and generate data for image classification.
• Combined CNN and LSTM for spatio-temporal feature extraction.
• Investigated the use of reinforcement learning for visual signal classification.

Demo [https://github.com/ArghaSarker/COVID-19-Global-data-EDA-]

• Conducted data visualization and exploratory analysis on global COVID-19 data.
• Identified trends and correlations among variables using statistical methods.

• Generative AI with Large Language Models (LLMs)

• Email: [argha.sarker.93@gmail.com]
• LinkedIn: [https://www.linkedin.com/in/argha-sarker-cogsci/]