Feng HUANG

Meet Feng

1 following

Applied AI developer with 2+ years building LLM-assisted workflows, AI search and chatbot applications, recommendation systems, and automated data pipelines.

What I do best: turning messy business data into usable AI-powered tools, with evaluation frameworks and human-in-the-loop review so the output can actually be trusted.

EXPERIENCE

Senior Data Specialist (2026)
- Built AI-assisted workflows to extract, standardise and validate financial data from unstructured Excel and PDF sources.
- Applied LLM-assisted classification to map inconsistent financial categories into structured outputs with reviewable validation rules.
- Designed data-quality checks and exception handling to flag bad source formats, missing fields and mapping errors before they reached stakeholders.

Data Scientist, consulting (2023-2025)
- Built an AI product-search chatbot using LLM workflows, web-scraped training data, and prompt/retrieval iteration.
- Built a job recommender that scored candidate-job fit and improved recommendation relevance by 23%.
- Built SQL/Python pipelines for ingestion, processing, scoring and delivery.
- Developed predictive models and used feature analysis to explain model behaviour for operational decision support.

SELECTED PROJECTS
- AI Job Match Agent - local-first resume/JD matching, scoring, tailored material generation and application tracking with human review.
- Fraud Case Triage Agent - hybrid architecture combining LLM reasoning, deterministic validation tools, risk scoring and explainable human-in-the-loop review.
- Recommendation System - pipeline over 5M+ behavioural records using candidate generation, LightGBM ranking and top-K evaluation.
- Credit scoring and NLP sentiment pipelines using LightGBM/XGBoost and spaCy/NLTK.

STACK
Python, SQL, pandas, data pipelines, APIs, web scraping, structured outputs, data validation | LLM workflows, RAG concepts, prompt engineering, AI agents, evaluation frameworks | scikit-learn, LightGBM, XGBoost, spaCy, NLTK, recommendation and ranking models | Streamlit, AWS S3/EC2, Git, Docker, Tableau, Power BI

EDUCATION
Master of Management in Artificial Intelligence (2026)
BSc Mathematics and Statistics

How I work: I start with a small, verifiable first milestone on your real data, so you can judge the quality before committing to the full scope. I would rather tell you something is out of scope than overpromise.