Hello, I’m

Ashiq Sazid

Founder @ CogniLynch · AI/ML Backend Developer @ Moodifai

I'm Md. Ashiq Ul Islam Sajid, author of 13 published research papers (6 first-author), with 4 under review. I build ML systems, contribute to open source, and target Q1 journals, USENIX, and NDSS. My goal is to become a skilled systems architect.
Currently,collaborating with Dr. Md. Alamgir Hossain (University of Southern Queensland) on FedGrid-RL: Attack Resilient Federated Reinforcement Learning for Household Battery Coordination in Queensland and working on two Q1 Journal.

13Published papers

4Papers under review

2Active AI projects

Get in touch →Learn more

Google Scholar GitHub LinkedIn ORCID YouTube

Research Interests

LLM Security
Adversarial Machine Learning
Trustworthy & Explainable AI
RL for Resource-Constrained Edge Systems
Privacy-Preserving & Federated Learning
Medical Image Analysis
Multimodal Learning
Natural Language Processing

Ongoing Research

Target: ICLR

Quantum circuits with large language models

Inspired by Anthropic's new paper Discovering state-of-the-art quantum circuits with large language models: an LLM-driven framework that designs and optimizes quantum circuits, cutting computational cost and qubit requirements in seven of eight tasks and discovering a new architecture for modular inversion. Planning to work in this domain.

Target: ICML

System One models

Brainstorming after JevAgentHarness: Benchmarking, Deploying and Training: efficient agents that let smaller decision models handle routine choices instead of calling an LLM at every step (AgentDecisionBench, JevAgentHarness, AgentDecision-4B), while keeping task success high.

Target: USENIX

Batch Echoes · Timing Side Channels in LLM Inference Servers

Measuring token-length timing leakage in continuous-batching schedulers (vLLM, SGLang), with pilot experiments on RTX 4090 infrastructure.

Read about these ideas on the blog →

Selected Publications

IEEE FET'26Selected

BitRL: Reinforcement Learning with 1-bit Quantized Language Models for Resource-Constrained Edge Deployment

2026 IEEE International Conference on Frontiers of Engineering and Emerging Technologies (FET) · Apr 2026

DOI

AHTBE'25Selected

MedViT-HoVer++ (ViT): A Unified Transformer-Guided Framework for Multitask Nucleus Segmentation, Classification, and Count Regression in Histopathology Images

2nd International Conference on Advancement in Healthcare Technology and Biomedical Engineering (AHTBE) · Presented Jun 2025

Code

AII'25Selected

Vertical AI for Kidney Stone Detection: Knowledge-Distilled CNNs with Student–Teacher Model for Ultrasound Imaging

5th International Conference on Applied Intelligence and Informatics (AII) · Presented Oct 2025

Paper

View all publications →