Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I am Songhui Zhen, an AI & NLP specialist based in Dublin, Ireland, holding an M.Sc. in AI & NLP from Dublin City University (top 5%). I spent more than a decade at the Shanghai Municipal Taxation Bureau, where I automated workflows, built knowledge bases, and contributed to the development of tax infrastructure software. My master's thesis involved collaboration with Walkers Global, using large language models to convert US tax code into executable Prolog predicates, emphasizing reliability and reusability. I am proficient in Python, PyTorch, and LLMs, and I am passionate about building robust, explainable AI systems.…I am Songhui Zhen, an AI & NLP specialist based in Dublin, Ireland, holding an M.Sc. in AI & NLP from Dublin City University (top 5%). I spent more than a decade at the Shanghai Municipal Taxation Bureau, where I automated workflows, built knowledge bases, and contributed to the development of tax infrastructure software. My master's thesis involved collaboration with Walkers Global, using large language models to convert US tax code into executable Prolog predicates, emphasizing reliability and reusability. I am proficient in Python, PyTorch, and LLMs, and I am passionate about building robust, explainable AI systems.

Songhui Zhen

Data Scientist, AI Engineer, AI Developer, +2





I am Songhui Zhen, an AI & NLP specialist based in Dublin, Ireland, holding an M.Sc. in AI & NLP from Dublin City University (top 5%). I spent more than a decade at the Shanghai Municipal Taxation Bureau, where I automated workflows, built knowledge bases, and contributed to the development of tax infrastructure software. My master's thesis involved collaboration with Walkers Global, using large language models to convert US tax code into executable Prolog predicates, emphasizing reliability and reusability. I am proficient in Python, PyTorch, and LLMs, and I am passionate about building robust, explainable AI systems.…I am Songhui Zhen, an AI & NLP specialist based in Dublin, Ireland, holding an M.Sc. in AI & NLP from Dublin City University (top 5%). I spent more than a decade at the Shanghai Municipal Taxation Bureau, where I automated workflows, built knowledge bases, and contributed to the development of tax infrastructure software. My master's thesis involved collaboration with Walkers Global, using large language models to convert US tax code into executable Prolog predicates, emphasizing reliability and reusability. I am proficient in Python, PyTorch, and LLMs, and I am passionate about building robust, explainable AI systems.

Available to hire

I am Songhui Zhen, an AI & NLP specialist based in Dublin, Ireland, holding an M.Sc. in AI & NLP from Dublin City University (top 5%). I spent more than a decade at the Shanghai Municipal Taxation Bureau, where I automated workflows, built knowledge bases, and contributed to the development of tax infrastructure software.

My master’s thesis involved collaboration with Walkers Global, using large language models to convert US tax code into executable Prolog predicates, emphasizing reliability and reusability. I am proficient in Python, PyTorch, and LLMs, and I am passionate about building robust, explainable AI systems.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Intermediate

Intermediate

Intermediate

Intermediate

Language

English

Fluent

Work Experience

Senior Specialist / Tax Officer at Shanghai Municipal Taxation Bureau

August 1, 2011 - August 1, 2024

Appointed to the national task force for the Golden Tax Project (Phases III & IV), architecting technical and data frameworks for China’s national tax infrastructure. Provided strategic oversight by reviewing vendor-delivered schemes to ensure high-concurrency reliability and scalability. Developed a Python-based RPA tool to automate cross-system data entry and designed a SQL-based full-text retrieval system for internal tax counseling. Awarded for excellence in municipal tax service delivery.

Masters Practicum Project: Neuro-Symbolic Legal Reasoning at Walkers Global

November 1, 2024 - August 1, 2025

Architected a multi-stage Neuro-Symbolic pipeline to automate the conversion of US Tax Code (SARA) into executable Prolog predicates, achieving 64% end-to-end accuracy. Developed a deterministic reasoning engine using Python, SWI-Prolog and LLMs (e.g., GPT-4o, Gemini 2.5 Pro) that optimizes cost by replacing redundant LLM calls with a verified symbolic knowledge base, ensuring consistent, single-source-of-truth answers. Engineered an automated feedback and self-correction loop boosting syntactic correctness to over 90% and ensuring 100% verifiable outcomes for high-stakes legal automation.

Research Intern at Huawei Ireland Research Centre – IOM Lab

July 1, 2025 - December 1, 2025

Evaluated various open-source autonomous networking AI frameworks, focusing on 'trustworthiness' and explainability gaps that hinder large-scale deployment in telecoms. Authored technical insight reports on SOTA AI techniques in autonomous networking, breaking down complex concepts into technical assessments and strategic implications for both technical and business stakeholders.