全职 Research Associate- Assistant (Ref. No.: AiDLab-P1.4-R106-107) 工作, 薪水 up to HKD 34,503.1, Laboratory For Artificial Intelligence In Design Limited Hong Kong 公司招聘中 - Ricebowl

LABORATORY FOR ARTIFICIAL INTELLIGENCE IN DESIGN LIMITED

Research Associate / Research Assistant (Several posts)

The appointees will assist the project leader in the research project – “ANTLER I”.

Post No. 1

You will be part of a dynamic Al and fashion team responsible for Model Training, Active Learning, and System Architecture. You will design and train the core Al models that detect errors and diagnose root causes. Working with clean datasets provided by the Data Engineer and ground-truth labels from our Embroidery Expert, you will build a hierarchical classification pipeline that is interpretable, efficient, and accurate even with limited initial training data.

Key responsibilities

• Model Training: Build a multi-stage pipeline. Stage 1 detects if an error occurred (GOOD vs. ERROR). Stage 2 classifies the specific type of error (A-G). Stage 3 determines the root cause.
• Active Learning: Implement systems that identify which video frames the AI is most "unsure" about, and automatically send those specific frames to our Embroidery Expert for manual labeling to improve the model efficiently.
• Model Evaluation: Use techniques like cross-validation and SMOTE (for class imbalance) to ensure models are accurate. Use SHAP values to explain why the model made a decision.
• Experiment Tracking: Run structured experiments to compare different algorithms (e.g., XGBoost vs. LightGBM vs. simple Neural Nets) and track them using tools like MLflow or Weights & Biases.
• Future Roadmapping: As our dataset grows, prototype more advanced deep learning architectures (like Temporal Convolutional Networks or LSTMs for sequence modeling).

Qualifications

For the Research Associate/ Assistant post, applicant should:

• 3+ years of experience developing Machine Learning models.
• Deep understanding of supervised learning (classification/regression) and how to handle real-world issues like imbalanced datasets and small sample sizes.
• Proficiency in Python and ML libraries (scikit-learn, XGBoost/LightGBM).
• Rigorous approach to model evaluation (preventing data leakage, proper train/test/holdout splits).
• Ability to build interpretable models (we need to know why the AI flagged an error, not just that it found one).

Candidates with less qualification and experience may be considered at the level of Research Assistant.

For the Research Assistant post, applicants should have a Bachelor of Science degree or Master of Arts degree in Computer Science or other related disciplines.

• Experience with PyTorch or TensorFlow for deep learning.
• Background in time-series or sequential modeling (LSTMs, Transformers).
• Experience building Active Learning pipelines.
• Familiarity working alongside domain experts to translate human knowledge into ML rules.

Post No. 2

Duties

You will be part of a dynamic Al and fashion team responsible for domain knowledge, data labelling, and system testing. An embroidery practitioner is needed to be the "ground truth" for our Al. You do not need to know how to code; you need to know what good (and bad) embroidery looks like. You will watch videos of people stitching, use our software tools to tag where they made mistakes, and test our Al to see if its feedback makes sense to a real human stitcher.

Key responsibilities

• Video Annotation: Watch embroidery sessions on a computer and use labelling software (like Label Studio) to tag specific errors (e.g., "stitch too long," "bad needle angle," "uneven tension").
• Root Cause Analysis: Identify why the mistake happened based on the video (e.g., "the student's grip was too loose").
• Rule & Taxonomy Creation: Help our technical team write clear definitions for different types of embroidery mistakes so everyone labels them consistently.
• AI Testing: Review the feedback our Al generates and grade it. (e.g., "Is this what a real teacher would tell the student?").

Qualifications

For the Research Associate/ Assistant post, applicant should:

• 2+ years of hands-on experience with hand embroidery (Tambour/Luneville embroidery is a massive plus, but advanced general embroidery experience works if you are willing to learn).
• A very sharp eye for detail you can easily spot when a stitch is slightly off and know why it happened.
• Comfortable using computers for long periods. You don't need to be a programmer, but you must be comfortable learning new web-based software tools (like video labelling platforms).
• Patient and methodical; data labelling requires focusing on repetitive tasks while maintaining high accuracy.
• Good communication skills to explain technical embroidery concepts to software engineers.

Candidates with less qualification and experience may be considered at the level of Research Assistant.

For the Research Assistant post, applicants should have a degree in Textiles or Fashion disciplines or other related disciplines, with demonstratable experience teaching embroidery or assessing students' work and previous experience with QA (Quality Assurance) or any kind of data entry/labelling.

Post No. 3

Duties

You will be part of a dynamic AI and fashion team responsible for Sensor Fusion, Feature Extraction, and Data Engineering. You will build the bridge between raw recording sessions and our Machine Learning models. You will take messy, multi-modal data (multiple camera angles, 100Hz IMU motion sensors, and MediaPipe hand-tracking data), synchronize it perfectly, and extract the mathematical features our ML team needs to train the AI.

Key responsibilities

• Data Synchronization: Ingest video and sensor streams and ensure they are perfectly synced using timestamps, handling dropped frames or sensor drift.
• Feature Extraction: Write Python scripts to extract specific metrics from the raw data. Examples include calculating stitch length from video, extracting needle angles from IMU gyroscope data, and measuring wrist angles using MediaPipe.
• Data Augmentation: Artificially expand our dataset by writing scripts to rotate videos, change lighting, or add safe noise to sensor data.
• Data Quality Control: Build automated tests to ensure the extracted features are robust against different lighting, fabrics, and camera angles before passing them to the ML team.
• Dataset Management: Maintain organized, version-controlled datasets (JSON, HDF5, Parquet) ready for model training.

Qualifications

For the Research Associate/ Assistant post, applicant should:

• 2+ years of experience in Data Engineering, Data Science, or Scientific Computing.
• Strong Python skills, specifically with libraries like pandas, numpy, and OpenCV.
• Experience processing time-series data (e.g., sensor data, audio, or financial ticks) and handling synchronization issues.
• Experience with Git version control and building reproducible data pipelines.
• A systematic, detail-oriented approach to debugging messy real-world data.

Candidates with less qualification and experience may be considered at the level of Research Assistant.

For the Research Assistant post, applicants should have a Bachelor of Science degree or Master of Arts degree in Computer Science or other related disciplines.

• Experience with Computer Vision tools (YOLO, MediaPipe, OpenCV).
• Background working with IMU (accelerometer/gyroscope) or biomechanical data.
• Familiarity with data augmentation libraries (like albumentations).

Applicants are invited to contact Anne Toomey at ************* or Zowie Broach at ************* for further information.

Please return the completed application form, together with a detailed curriculum vitae, to the AiDLab by email to ************* Application form can be downloaded from AiDLab offical website.

Remuneration

A highly competitive remuneration package will be offered. Applicants should state their current and expected salary in the application.

Job Types: Full-time, Part-time, Contract

Pay: $15,514.24 - $34,503.09 per month

Benefits:

• Dental insurance
• Employee pension
• Maternity leave
• Medical Insurance
• Paid sick time
• Professional development

Work Location: In person