Henry C G Baker
About Me
I am a Research Engineer at the Hertie Data Science Lab in Berlin, where I develop lab infrastructure and support computational research. I hold a dual degree in Public Policy and Data Science from the Hertie School of Governance, with research interests in computational social science, the societal and environmental dynamics of ML & AI applications, and green AI.
My background spans East Asia, Europe, and Sub-Saharan Africa, with experience in policy research, product management, and applied data science. Once upon a time before that even, I was a Modern History undergraduate at the University of Oxford - which I like to think continues to inform my work and thinking today.
Research Interests
- Sustainable Computing & Green AI: energy efficiency of ML systems, computational sustainability
- Applied ML: Deep Learning, NLP, Computer Vision, Causal ML - applied to policy-relevant problems
- Computational Social Science: Bayesian hierarchical modelling, forecasting, geospatial analysis, causal inference
Affiliations
- Hertie Data Science Lab (Berlin) β Research Engineer
- Weizenbaum Institute (Berlin) β Associate Researcher, Digitalization, Sustainability, and Participation
- Open Data Institute (London) β Research Fellow, Data-centric AI project
Technical Skills
- Languages: Python, R, SQL, Bash, Stan, LaTeX
- Deep Learning: PyTorch, TensorFlow, JAX, HuggingFace libs
- Data Analysis: Scikit-Learn, Pandas, NumPy, Matplotlib
- MLOps: Docker, MLflow, Heroku, HF Spaces
- Deployment \& Inference: Flask, vLLM, TensorRT
- UI & Web: HTML/CSS, Streamlit, Gradio
Project Management: JIRA, Confluence
Languages
- English β Native
- Japanese β β½ζ¬θͺθ½βΌθ©¦ι¨βΌη΄ (Business proficient)
- French β Conversant
Teaching
Teaching Associate on the Hertie School of Governanceβs MSc Data Science programme:
- Mathematics for Data Science (Lead TA, β24)
- Data Structures & Algorithms (Lead TA, β25)
- Deep Learning (TA β25)
Research Projects
- LLM Energy Efficiency Measurement: Benchmarking transformer model energy consumption
- ML-Strom: Machine learning for electricity grid analysis (with L. Hirth & L. Kaack)
- UK Climate Attitudes: Bayesian hierarchical IRT model to extract latent public sentiment from proprietary longitudinal polling data
- DiD Analysis: Difference-in-Differences causal inference studies (with M. Kayser)
Software
- LLenergyMeasure: Python framework for benchmarking energy consumption, throughput, and FLOPs in LLM inference across multiple backends (PyTorch, vLLM, TensorRT)
Powered by Jekyll and Minimal Light theme.