Hello, I’m
I'm Md. Ashiq Ul Islam Sajid, have published 13 research papers at international conferences and in Q1 journals, with 4 more currently under review.
Besides research, I build machine learning systems both independently and collaboratively, contribute to open-source projects on GitHub, and am actively working on Q1 journal and an A* conference.
Targeting USENIX & NDSS; becoming a good system architect is my main priority now.
Building and deploying AI backend services model integration, REST API development, and scalable inference pipelines for production systems.
USA-based startup and research company.
TheraMuse; an adaptive music therapy platform for children with Down syndrome, individuals with ADHD, and patients with dementia. Built around 32 clinically informed parameters with Q-Learning and Linear Thompson Sampling for real-time personalization. Recently secured $3 million in VC funding, supervised by Dr. Mustak Ibn Ayub at the University of Oxford.
Studying membership inference attacks, shadow model training, and differential privacy (DP-SGD / Opacus) as defenses in federated learning settings.
Novel view synthesis for Visual Place Recognition — analyzing how hallucinated pixels from diffusion inpainting affect recognition performance across real-world datasets.
Applying machine learning to real-world security challenges — including IoT intrusion detection, SDN-based threat detection, and resource-efficient multi-class threat classification in constrained environments.
13 accepted, 4 under review; first author on 6, co-first author on all
LoRA fine-tuning workflow and experiments built around Qwen2.5 models.
View Project →Transformer-guided framework for multitask nucleus segmentation, classification, and count regression in histopathology images.
View Project →A small demo code of Military-grade YOLOv8 pipeline built with Bangladesh Army collaboration, optimized for blurry CCTV inference.
View Project →FastAPI backend for a healthcare platform: auth, medical document uploads, secure storage, and AI-assisted workflows for prescription/image analysis, doctor recommendation, and RAG-style support.
View Project →Python, Django, Flask, FastAPI, Git, Linux.
TensorFlow, PyTorch, Keras, scikit-learn, Hugging Face Transformers.
Adversarial robustness, attack analysis, threat modeling, red teaming, secure model evaluation, and resilience testing for AI systems.
RL systems on 1-bit quantized language models for resource-constrained edge deployment.
AWS, Azure, Kafka, Docker, MySQL, MongoDB, SQLite.
Solidity, JavaScript, Truffle, Ethereum, Hyperledger Fabric, Ganache, MetaMask, Web3.js, ERC-721.
LLM fine-tuning, RAG pipelines, prompt engineering, LangChain, multilingual NLP (Bangla/English).
Image segmentation & classification, U-Net / Vision Transformers, OpenCV, MONAI, knowledge distillation for diagnostic models.
Next.js, React, TypeScript, Tailwind CSS, Spring Boot, REST API design.
Docker, CI/CD, model quantization, ONNX, spot-instance training, Vercel / Render / Cloudflare deployment.
Open to AI/ML backend roles, research collaborations, and product-focused ML work.