Senior AI Research Scientist

Senior AI Research Scientist(6240)

San Jose, CA

TSMC seeks a Senior AI Research Scientist for its AI4BI Center to develop advanced AI-enabled analytics techniques for business intelligence applications highly relevant to TSMC.

• Create and manage advanced machine learning, deep learning, LLM, foundation models, time‑series modelling, and/or reinforcement learning algorithms to develop data‑driven solutions for complex business problems.
• Analyze and interpret complicated data sets to provide actionable insights that aid in enhancing decision‑making processes for business stakeholders.
• Work with cross‑functional and international teams to identify and prioritize data‑driven opportunities, including the development and deployment of predictive models to drive business outcomes.
• Lead teams of junior data scientists and data engineers to develop efficient procedures, follow best practices, and provide implementation guidance (e.g., code reviews, demos) for gathering, retaining, and scrutinizing data.
• Consistently assess and refine model performance to ensure precision and dependability.
• Communicate insights and recommendations via internal reports to technical and non‑technical stakeholders.
• Design and implement the latest AI, machine learning, and data science tools and techniques to enhance existing processes.
• Produce publications in leading AI conference and/or journal venues.

• Ph.D. in computer science, information systems, information science, statistics, or related AI or machine learning field.
• At least 7‑10 years of significant AI, machine learning, and data science‑related working and research experience.
• Experience gathering requirements from business stakeholders and developing technical machine learning designs for implementation.
• Strong practical experience with data wrangling, pre‑processing, and extraction from structured and unstructured data sources.
• Strong conceptual and practical knowledge and experience of classical machine learning algorithms and learning paradigms, including supervised learning and unsupervised learning.
• Strong skills in deep learning implementation, including data encodings, processing units, and learning paradigms.
• Proficiency in developing novel machine learning or deep learning algorithms based on unique dataset characteristics and business requirements.
• Strong practical network science skills and experience.
• Experience with text analytics for named entity recognition, sentiment analysis, and topic modelling.
• Hands‑on research experience with time series modeling and forecasting with statistical, machine learning, deep learning‑based, and/or foundation model (Times
  
  FM) approaches.
• Hands‑on research experience with reinforcement learning approaches, including value‑based, policy‑based, actor‑critic, and/or multi‑objective.
• Hands‑on research experience, including literature review, model design, implementation, evaluation, and publication.
• Experience fine‑tuning LLMs and other foundation models with strategies such as low rank adaptation, few shot learning, and others.
• Experience with prompt engineering on LLMs and other foundation models with techniques such as reinforcement learning (e.g., Direct Preference Optimization (DPO), Proximal Policy Optimization (PPO)), prompt tuning, and others.
• Knowledge about deploying models into user interface tools for dashboarding (e.g., Power
  
  BI), Python frameworks for rapid prototyping and interaction (e.g., Streamlit, Dash), and/or JavaScript frameworks (e.g., React).
• Must be willing to travel to Taiwan for at least 1‑3 months each year for training, team building, and project coordination.

• Knowledge on processing text in financial, accounting, and market analysis reports.
• Multi‑lingual text analysis using packages like Stanza, Polyglot, or Textflint.
• Knowledge of graph embedding techniques such as graph convolutional networks, graph attention networks, and graph transformers using packages such as stellar graph, PyG, or Deep Graph Library.
• Understanding of neural information retrieval methods, including deep structured semantic models, entity resolution, and retrieval augmented…