Hao Liang - AI/ML Engineer

Hey there 👋

I'm Hao Liang 梁昊

AI/ML Engineer

Specializing in production LLM systems, cloud-native data platforms, and MLOps for large-scale environments. Experienced in end-to-end system design spanning modeling, optimization, governance, and deployment.

LinkedIn Google Scholar

Core Technical Specializations

Agentic AI & LLM Systems

Architecting production-grade knowledge systems and autonomous workflows using RAG pipelines, prompt engineering, and LLM orchestration.

Distributed Data Engineering

Designing scalable cloud-native architectures and high-throughput ETL/ELT pipelines to manage complex, high-dimensional datasets.

Algorithmic Optimization & Decision Logic

Translating multi-dimensional constraints into scalable decision-making logic using graph-based algorithms and mathematical optimization frameworks.

Production MLOps & Cloud Infrastructure

Managing the end-to-end machine learning lifecycle through containerized microservices, Infrastructure as Code, and automated CI/CD pipelines.

Advanced Predictive Modeling

Developing statistical and neural architectures for high-fidelity pattern recognition, time-series forecasting, and automated data synthesis.

Spook Distribution

"In God we trust, all others must bring data."

Developer Reality

"It works on my machine... but I don't live on my machine."

Professional Experience

Roche

Data Engineer

Mar 2024 – Present Branchburg, NJ

Design and deliver end-to-end AI/ML, data engineering, and optimization systems supporting Roche’s R&D and product pipelines

AI & LLM Engineering and Optimization

Architect production-grade LLM knowledge systems by building RAG pipelines and benchmarking NLP models (LDA, BERT, BART) to automate R&D data synthesis
Integrate evaluation frameworks (Ragas) to quantify retrieval accuracy and generation quality, ensuring high-fidelity outputs for sensitive scientific workflows
Build OCR and embedding-based parsing pipelines to convert compliance documents into structured data, automating form-filling and reducing manual effort
Apply graph-based optimization algorithms (e.g., Dijkstra's) to automate combinatorial oligonucleotide design, translating scientific constraints into scalable decision logic
Optimize manufacturing logistics and reduce line failures using Optimization to balance inventory and production targets

Data Architecture and Governance

Engineer Data Vault 2.0 on Snowflake, integrating SAP, MES, and SQL Server sources via Qlik, Talend, and dbt to establish a governed, analytics-ready data foundation
Build a data governance stack integrating Monte Carlo, Collibra, and Immuta to enable observability, metadata management, and secure access across AI workloads

Cloud Infrastructure and MLOps

Modernize production infrastructure by migrating 10+ projects to AWS ECS/Fargate using GitLab CI/CD and Infrastructure as Code (IaC)

Analytics Platforms & BI Automation

Pioneer the first Posit Connect Server deployment to host Streamlit applications, replacing manual workflows with containerized, production-ready analytics environments
Lead multi-site Tableau rollouts by automating ingestion from Monday.com and Smartsheet into Snowflake, improving operational visibility and reporting consistency

Cisco

Data Science Consultant

Sep – Dec 2023 San Jose, CA

Empowered supply chain teams with an end-to-end analytics platform to detect market shifts and potential disruption risks

Built ELT pipelines to aggregate semiconductor news and commodity price data via web scraping (Beautiful Soup) and external APIs, creating a unified signal layer for downstream analysis
Leveraged fine-tuned prompt engineering with LLM APIs to automate news summarization and topic modeling
Distilled risk signals using NLP (NER, sentiment, lemmatization) in NLTK and spaCy, improving intent and risk models
Developed Tableau dashboards to track emerging risks and support data-driven decisions for supply chain stakeholders

H-E-B

Data Scientist Intern

Jun – Aug 2023 Austin, TX

Applied ML, AI, and optimization to solve large-scale supply chain challenges, improving efficiency and driving innovation

Engineered 100M+ record datasets with PySpark SQL, building scalable ETL pipelines and data-quality checks for forecasting and optimization
Developed demand forecasting models using time-series methods (e.g., SARIMAX) and deep learning architectures (e.g., GRU, LSTM), incorporating feature-driven signals to capture demand dynamics
Designed an inventory simulation framework to evaluate policy performance under demand uncertainty and operational constraints; applied Hyperopt to tune forecasting and inventory parameters
Optimized inventory strategies with Google OR-Tools and heuristics, producing actionable supply-chain recommendations

Sabre

Data Science Consultant

Sep 2022 – Jun 2023 Dallas, TX

Developed and deployed statistical and ML solutions to support dynamic pricing and market strategy in the airline industry

Built airline price elasticity models using Poisson Regression, XGBoost, Random Forest, and LSTM, improving demand sensitivity estimation across markets and fare granularities to inform revenue and market entry strategies
Performed market segmentation using K-means, Agglomerative Clustering, and DBSCAN, enabling personalized prediction and differentiated pricing strategies across customer and route segments
Architected automated ML pipelines using YAML, Docker, and Python-based UIs for model training and real-time inference, accelerating experimentation cycles and enabling production-grade predictions

Refinitiv (London Stock Exchange Group)

Data Scientist Intern

Apr – Jul 2021 Beijing, China

Delivered data-driven insights on user behavior and product engagement to support analytics and platform strategy for Eikon

Design and run A/B experiments on Eikon webpage layouts, driving a 16% increase in Daily Active Users (DAU)
Build dashboards for bond credit rating monitoring using Python (Bokeh), reaching 5,000+ views in the first month
Analyze subscriber behavior and retention with R to inform product and marketing strategy
Authored and delivered Python tutorials on Eikon API usage; trained 1,000+ users in programmatic data access and analysis

Minsheng Securities

Data Analyst Intern

Aug – Oct 2020 Beijing, China

Built data pipelines, analytical models, and forecasting tools to inform trading and client-profitability analysis

Built and maintained ETL pipelines for Private Placement transaction data using SQL, extracting key trading and client metrics; performed feature engineering and exploratory data analysis in Python to prepare modeling-ready datasets
Developed time-series forecasting models (e.g., ARIMA) in R (forecast) to predict Private Placement trading volume, achieving a MAPE of approximately 15% and supporting short-term market activity planning
Analyzed client profitability and portfolio relationships using factor-exposure analysis to identify concentration risk
Translated analytical findings into structured insights for stakeholders, informing trading and investment decisions

Dongxing Securities

Risk Data Analyst Intern

Jun – Aug 2020 Beijing, China

Applied quantitative modeling and analytics to assess portfolio risk, support equity research, and inform client investment insights

Quantified portfolio risk by computing and comparing VaR and ES, enabling robust cross-method risk assessment
Developed risk analysis workflows to evaluate portfolio sensitivity under different market scenarios, synthesizing results into structured risk assessment reports for internal teams and external clients
Performed equity valuation and factor analysis, implementing Discounted Cash Flow (DCF) models and estimating equity beta via linear regression in R to support fundamental research and risk-adjusted return analysis
Built visualizations to analyze stock-price dynamics and key indicators (RSI, MACD) for equity-research insight

Research & Insights

Publications & Conference Proceedings

Liu M., Zhang W., & Liang H.

Value effect of AI innovation zones: Green premium and cost reduction pathways in environmental disclosure

Research in International Business and Finance

2026

Liang, H. & Hernke M.

Quantifying the Environmental Impact of Electric Vehicle Adoption

International Conference on Frontiers of Energy and Environment Engineering (CFEEE)

2024

Li, Y., Huynh, L., Xu, Y., & Liang, H.

The Forecast Ability of a Belief-Based Momentum Indicator in Full-Day, Daytime, and Nighttime Volatilities of Chinese Oil Futures

Energy Economics

2023

Journal Peer Reviewer

Journal of Innovation & Knowledge

Finance Research Letters

International Review of Financial Analysis

Research in International Business and Finance

International Review of Economics and Finance

Evaluation Review

International Finance

Review of Accounting and Finance

Humanities and Social Sciences Communications

Research Experience

University of Wisconsin – Madison

Research Assistant (Prof. Michael Hernke)

Sep 2019 – Dec 2021 Madison, WI

Wisconsin School of Business

Constructed a break-even analytical framework to investigate the Jevons’ Paradox threshold, quantifying the energy-efficiency gains required to achieve net environmental savings in alternative energy vehicle markets
Designed and executed a multi-source data integration architecture, harmonizing high-dimensional datasets from state and federal agencies to establish normalized energy features for longitudinal econometric analysis
Utilized input-output analysis to model ecological footprints, establishing a quantitative framework for assessing cross-sectoral environmental impacts of emerging transportation technologies
Authored peer-reviewed research on the rebound effect in EV adoption, presented at international conferences

University of Wisconsin – Madison

Research Assistant (Prof. Margie Rosenberg)

Sep 2019 – May 2020 Madison, WI

Wisconsin School of Business

Implemented a Difference-in-Differences (DiD) model to quantify the causal effect of economic downturns on population-level health indicators across demographic cohorts
Engineered a harmonized longitudinal database from 20+ public health surveys (including BRFSS and NHANES) using survey-weighted adjustments and cell-based weighting
Authored the research proposal defining the technical methodology and led the initial data-engineering architecture

Agentic AI & LLM Systems

Distributed Data Engineering

Algorithmic Optimization & Decision Logic

Production MLOps & Cloud Infrastructure

Advanced Predictive Modeling

Roche

AI & LLM Engineering and Optimization

Data Architecture and Governance

Cloud Infrastructure and MLOps

Analytics Platforms & BI Automation

Cisco

H-E-B

Sabre

Refinitiv (London Stock Exchange Group)

Minsheng Securities

Dongxing Securities

Northwestern University

University of Wisconsin – Madison

Value effect of AI innovation zones: Green premium and cost reduction pathways in environmental disclosure

Quantifying the Environmental Impact of Electric Vehicle Adoption

The Forecast Ability of a Belief-Based Momentum Indicator in Full-Day, Daytime, and Nighttime Volatilities of Chinese Oil Futures

University of Wisconsin – Madison

University of Wisconsin – Madison

AI, ML & LLM Systems

Programming Languages

Data Engineering & Platforms

Cloud & DevOps

Web & Application Frameworks

Data Visualization & BI

Certificate in Sustainable Investing

Financial Risk Manager (FRM)