Job Description

Research and develop LLM‑based methods for clinical information extraction and ICD‑10/ICD‑11 coding.

• Conduct literature reviews on state‑of‑the‑art approaches in clinical NLP and coding (prompting, RAG, fine‑tuning, hybrid rule+LLM).
• Implement and experiment with prompting strategies (zero/few‑shot, chain‑of‑thought, self‑consistency) and tool/function calling.
• Build NLP pipelines including de‑identification, section segmentation, NER/RE (clinical entities and relations), normalization, and code assignment.
• Compare multiple modeling approaches (prompt‑only, RAG, fine‑tuned LLMs) to identify the best strategy for accuracy, robustness, and explainability.
• Analyze model limitations (hallucination, specificity errors, ambiguous terms) and propose improvements or alternative modeling strategies.

Apply LLM techniques (Prompting, RAG, Fine‑tuning) to real clinical texts for ICD‑10/ICD‑11 assignment.

• Fine‑tune or adapt pre‑trained models (e.g., LLaMA, Mistral, GPT‑style, ClinicalBERT/biomed models for components) with LoRA/QLoRA where permissible.
• Design knowledge‑retrieval components (indexes of ICD catalogs, coding guidelines, synonyms) to support evidence‑grounded code suggestions.
• Evaluate robustness under real‑world variations: misspellings, abbreviations, negation/uncertainty, multi‑lingual notes, domain shift across departments.
• Develop inference pipelines optimized for different scenarios (interactive coding assistance vs. batch auto‑coding).
• Document and communicate model performance, limitations, compliance constraints, and deployment considerations in clinical settings.

Evaluate model performance and optimize for precision, recall, speed, and cost.

• Measure performance with clinical‑coding metrics: micro/macro‑F1 at code level, precision@k, hierarchical metrics (chapter/section), latency, throughput, and cost.
• Perform detailed error analysis.
• Tune hyperparameters and serving parameters (context length, temperature, top‑p, beam settings, retrieval top‑k) to improve quality and efficiency.

Prepare technical reports, present research outcomes, and engage in academic exchange with Taiwanese students.

• Write technical documents detailing methods, experimental setups, datasets/preprocessing, results, and analysis with code snippets/configurations.
• Prepare presentation slides and deliver periodic progress updates in research meetings or seminars.
• Present final results to the research group and participate in academic discussions with Taiwanese students.
• Maintain weekly research logs and progress reports for supervisors and IIPP documentation.
• Create reproducible project documentation (prompt templates, retrieval index build steps, evaluation scripts) so others can replicate and extend the work.

Preferred Intern Educational Level

Master’s student in Computer Science, Biomedical Engineering, Health/Biomedical Informatics, Data Science.

Skill sets or Qualities

Technical Skills

• Python; PyTorch/Transformers; experience with HuggingFace ecosystem.
• LLM techniques: prompt engineering, Retrieval‑Augmented Generation (vector indexing/reranking), fine‑tuning (LoRA/QLoRA), guardrails and evaluation.
• Clinical NLP components: de‑identification, section segmentation, NER/RE, terminology normalization, ontology mapping (ICD‑10/ICD‑11).
• Data skills: text cleaning, annotation/labeling, dataset versioning, secure handling of PHI and access control.

Personal Qualities

• Strong analytical thinking and attention to detail.
• Ability to work independently as well as collaboratively in a team.

Job Description

Research and develop algorithms for image recognition, object detection, and computer vision analysis.

• Conduct literature reviews on state‑of‑the‑art methods in computer vision (classification, detection, segmentation).
• Implement and experiment with CNNs, Vision Transformers, and detection models such as YOLO, Faster R‑CNN, or Mask R‑CNN.
• Build image‑processing pipelines including preprocessing, feature extraction, and evaluation.
• Compare multiple algorithms to identify the best approach for real‑world applications.
• Analyze model limitations and propose improvements or alternative modeling strategies.

Apply deep learning techniques (CNNs, Vision Transformers) to medical imaging, industrial inspection, or smart‑city systems.

• Fine‑tune pre‑trained models (e.g., ResNet, EfficientNet, ViT, Swin Transformer) on domain‑specific datasets such as X‑ray, CT‑scan, industrial defect images.
• Design appropriate data‑augmentation strategies tailored to each domain.
• Evaluate model robustness under real‑world variations such as lighting, noise, camera angle, and sensor quality.
• Develop inference pipelines optimized for real‑time or high‑throughput scenarios depending on application requirements.
• Document and communicate performance, limitations, and applicability for each domain.

Evaluate model performance and optimize for speed and accuracy.

• Measure model performance using metrics such as Accuracy, F1‑score, mAP, IoU, ROC‑AUC, latency, and throughput.
• Perform detailed error analysis to understand misclassifications or missed detections.
• Optimize neural networks using pruning, quantization, knowledge distillation, or ONNX/TensorRT acceleration.
• Tune hyperparameters (learning rate, batch size, optimizers, schedulers) to improve model performance.
• Test models on multiple hardware configurations (CPU/GPU) to refine inference speed and stability.

Prepare technical reports, present research outcomes, and engage in academic exchange with Taiwanese students.

• Write technical documents detailing algorithms, experimental setups, results, and analysis.
• Prepare presentation slides and deliver periodic progress updates in research meetings or seminars.
• Present final results to the research group and participate in academic discussions with Taiwanese students.
• Maintain weekly research logs and progress reports for supervisors and IIPP documentation.
• Create reproducible project documentation so other researchers can replicate and extend the work.

Preferred Intern Educational Level

Master’s student in Computer Science, Electrical and Electronic Engineering, Biomedical Engineering, Smart Agriculture, or related fields.

Skill sets or Qualities

Technical Skills

• Python, PyTorch, TensorFlow.
• Computer Vision: CNNs, Vision Transformers, image preprocessing.
• Skills in data processing, data labeling, and model evaluation.

Personal Qualities

• Strong analytical thinking and attention to detail.
• Ability to work independently as well as collaboratively in a team.