QS World University Rankings · Top 500
01 — Introduction
Dr. Ijazul Haq is an AI Research Engineer at Guangdong Casangels Biotechnology Co., Ltd. and recently completed a joint postdoc at South China University of Technology (QS #342). He holds a PhD in Cyberspace Security from Shanghai Jiao Tong University (QS #36), focused on NLP, generative AI, and low-resource language technology.
He leads a five-engineer team across the full AI lifecycle—data, training, evaluation, and deployment of assistants, chatbots, and agents—chiefly in healthcare and dentistry. He also holds an MS from Wuhan University (QS #165) and a BS in Software Engineering from the University of Malakand.
Public resources include NLPashto, PsBERT, PsOCR, and the COde dental dataset, released on Hugging Face, GitHub, PyPI, and Kaggle.
02 — Experience
Guangdong Casangels Biotechnology Co., Ltd. · Guangzhou, China
Lead a team of five AI engineers designing and deploying AI solutions across domains, with emphasis on healthcare and dentistry. Develop and customize large models for production use. Responsible for the complete lifecycle: data collection, preprocessing, training, fine-tuning, evaluation, and deployment, including AI assistants, chatbots, and autonomous agents.
South China University of Technology (QS #342) with Casangels
Completed a two-year joint postdoctoral fellowship. Research concentrated on generative AI, low-resource NLP, and computer vision, producing multiple papers in high-impact journals. Concurrently translated experimental findings into deployable healthcare AI systems, bridging academic research and industrial R&D.
University of Malakand · Pakistan
Taught undergraduate courses in Database Systems, Programming Fundamentals, Boolean Algebra, Web Engineering, and Machine Learning Basics. Evaluated students, conducted examinations, and supervised independent academic research.
CFTX · Shanghai, China
Preprocessed time-series financial data for models predicting cryptocurrency market trends. Built web platforms for data collection and annotation to support the data-engineering pipeline at a fintech firm during doctoral studies.
Wuhan University · Software Engineering Department, China
Supported undergraduate teaching during the final six months of the master’s program: lectures, laboratory sessions, and examination invigilation under faculty supervision.
03 — Education
China · QS World University Rankings #36 · Department of Cyberspace Security
Five-year doctorate applying NLP and generative AI to online social-network security: content analysis and detection of offensive behavior, including cyberbullying and harassment. Developed state-of-the-art models for low-resource languages and published multiple high-quality papers in this domain.
China · QS #165 · School of Computer Science
Software Engineering and Blockchain Technology. Research on blockchain applications in supply-chain management and prevention of counterfeit pharmaceuticals.
Pakistan · Software Engineering
Four-year degree with a strong theoretical foundation in software engineering and computer science, practical software development, and early independent research training.
Pakistan · Diploma of Associate Engineering in Electronics Technology
Three-year diploma covering electronic equipment, circuit design, and computer hardware—the systems foundation for subsequent work in software engineering and AI.
04 — Languages & technical stack
Technical practice covers the modern generative-AI stack as well as classical machine learning and production engineering.
AI methods. Machine learning, deep learning, NLP, computer vision, multimodal AI, generative AI, agentic AI, assistants, and chatbots. Training and fine-tuning; RAG; RLHF; quantization; PEFT; LoRA; PPO; DPO. Synthetic data generation, augmentation, and knowledge distillation. Benchmarking, evaluation, and explainable AI (SHAP, LIME).
Architectures. CNNs, RNNs, Transformers, large language models, vision-language models, and diffusion models. Embeddings (Word2Vec, FastText, GloVe) and vector databases (FAISS, Pinecone, Milvus).
Engineering. Python, PyTorch, TensorFlow, NumPy, Pandas, LangChain, LLaMA-Factory, vLLM, SGLang, and agent-orchestration frameworks. Model deployment, REST APIs, Azure, AWS, and web platforms in PHP, MySQL, HTML, CSS, and JavaScript.
05 — Awards & grants
Fully funded doctoral scholarship at Shanghai Jiao Tong University covering tuition, living expenses, accommodation subsidies, and medical insurance. Awarded September 2018.
Fully funded master’s scholarship at Wuhan University covering tuition, living expenses, accommodation allowance, and medical insurance. Awarded September 2016 by the China Scholarship Council.
06 — Selected publications
Peer-reviewed work in generative-model evaluation, low-resource Pashto NLP, multimodal dental AI, software analytics, and social-media safety. Venues include Nature Scientific Data, Expert Systems with Applications, Ain Shams Engineering Journal, Information and Software Technology, and Language Resources and Evaluation.
Haoxin Lv, Ijazul Haq, Jin Du, Jiaxin Ma, Binnian Zhu, Xiaobing Dang, Chaoan Liang, Ruxu Du, Yingjie Zhang, Muhammad Saqib. Scientific Data (Nature), 2026. doi:10.1038/s41597-026-07342-9
Kanwal Ahmed, Muhammad Asghar Khan, Ijazul Haq, Alanoud Al Mazroa, Syam M.S., Nisreen Innab, Masoud Alajmi, Hend Khalid Alkahtani. Expert Systems with Applications, 255, 124650, 2024. doi:10.1016/j.eswa.2024.124650
Ijazul Haq, Yingjie Zhang, Muhammad Saqib. Ain Shams Engineering Journal, 17(3), 104024, 2026. doi:10.1016/j.asej.2026.104024
Muhammad Saqib, Ibrar Ali Shah, Ijazul Haq, Muhammad Faisal Abrar, Sadaqat Jaan, Muhammad Faisal Nadeem. Information and Software Technology, 195, 108124, 2026. doi:10.1016/j.infsof.2026.108124
Ijazul Haq, Muhammad Saqib, Yingjie Zhang, Irfan Ali Khan. Multimedia Systems, 32, 146, 2026. doi:10.1007/s00530-025-02150-4
Ijazul Haq, Weidong Qiu, Jie Guo, Peng Tang. Speech Communication, 153, 102970, 2023. doi:10.1016/j.specom.2023.102970
Ijazul Haq, Weidong Qiu, Jie Guo, Peng Tang. PeerJ Computer Science, 9, e1617, 2023. doi:10.7717/peerj-cs.1617
Ijazul Haq, Yingjie Zhang, Intakhab Alam Qadri. Language Resources and Evaluation, 2025. doi:10.1007/s10579-025-09834-3
Muhammad Saqib, Ijazul Haq, Ibrar Ali Shah, Muhammad Faisal Abrar, Muhammad Faisal Nadeem, Sadaqat Jan. Journal of Software: Evolution and Process, 38(10), e70177, 2026. doi:10.1002/smr.70177
Irshad Ali Khan, Gul Zaman Khan, Yar Muhammad, Samreen Ihsan, Ijazul Haq. ICCK Transactions on Applied Intelligence and Cybernetics, 1, 5–35, 2026. doi:10.62762/TAIC.2025.965486
Ijazul Haq, Weidong Qiu, Jie Guo, Peng Tang. Proceedings of the 8th International Conference on Signal and Image Processing (ICSIP), 614–618, 2023. doi:10.1109/ICSIP57908.2023.10270845
Ijazul Haq, Weidong Qiu, Jie Guo, Peng Tang. International Journal of Advanced Computer Science and Applications, 14(6), 1344–1352, 2023. doi:10.14569/IJACSA.2023.01406142
Ijazul Haq, Olivier Muselemu Esuka. International Journal of Computer Applications, 180(25), 2018. doi:10.5120/ijca2018916579
07 — Featured projects
An agent that automates implant planning from a single CBCT scan. The system identifies missing teeth and anatomical locations, then generates patient-specific 3D implant designs (width, height, diameters, angulation), replacing a workflow that otherwise requires specialized oral-scanning hardware and hours to days of clinical analysis.
LLM assistant inside a respiratory-rehabilitation product that pairs a sensor device with a mobile app. The assistant teaches device use and exercises, analyzes historical breathing performance, and produces personalized training plans, difficulty recommendations, and feedback.
First large-scale public Pashto OCR dataset: one million synthetically generated images annotated at token, line, and document levels with bounding boxes, covering 1,000 font families and extensive layout variation. Addresses OCR for Pashto and other low-resource Perso-Arabic scripts.
huggingface.co/datasets/zirak-ai/PashtoOCRClinical assistant that consumes multimodal patient data—histories, imaging, laboratory tests—to support treatment selection, prescription assistance, personalized plans, and oncologist decision-making for NPC.
Led construction of a dental AI dataset from 9,000 checkups of 4,800 patients (ages 10–90): 50,000 intraoral RGB images, 9,000 radiographs, and structured clinical text. Extracted 120+ conditions. Built an LMM assistant that accepts X-rays, CBCT, intraoral photos, and text to classify anomalies and recommend diagnosis, treatment, and follow-up.
huggingface.co/datasets/zirak-ai/COdeDiffusion-based (DALL·E 2) tool that predicts post-treatment oral appearance from clinical photographs. Two modes: clinician-highlighted anomaly regions, or automatic region detection before synthesis, giving patients and dentists a visual forecast of outcomes.
Python library on PyPI with datasets and models for cleaning, tokenization, segmentation, POS tagging, NER, offensive-language detection, embeddings, and Pashto BERT—foundational infrastructure for a previously underserved language.
github.com/zirak-ai/nlpashtoPublic Pashto offensive-language dataset (Kaggle) used to fine-tune PsBERT for sentiment and toxicity classification, supplying content-moderation resources for a low-resource language.
Monolingual BERT trained from scratch on a purpose-collected Pashto corpus and released on Hugging Face. Widely reused as a foundation model for downstream Pashto NLP tasks.
huggingface.co/zirak-ai/pashto-bert-v109 — Correspondence
For academic collaboration, invited talks, research partnerships, and professional inquiries, write to the address below. This site is the canonical public curriculum vitae for Dr. Ijazul Haq.
