Primary Goal: This repository aims to pull links together, and organize a format that brings faculty, students, and researchers closer together.
Secondary Goal: Deepen theory (probabilistic graphical models, RL, ML theory) and extend it into scalable, reproducible, and trustworthy ML systems. This repo tracks coursework prep, research scaffolding, and reproducible code toward CMU’s MS in Machine Learning – Advanced Study and a pathway into research with 1) ICSD or 2) Safe AI Lab.
-
CMU MS in Machine Learning – Advanced Study (MSML-AS):
https://ml.cmu.edu/academics/primary-ms-machine-learning-advanced-study-masters -
ML Department (SCS) – Program Overview & Curriculum:
https://www.ml.cmu.edu/
https://ml.cmu.edu/academics/machine-learning-masters-curriculum -
Master’s Student Handbook (latest posted):
https://ml.cmu.edu/current-students/index -
Institute for Complex Social Dynamics (ICSD):
https://www.cmu.edu/dietrich/social-dynamics/ -
Safe AI Lab @ CMU:
https://safeai-lab.github.io/
https://www.ri.cmu.edu/robotics-groups/cmu-safe-ai-lab/ -
Faculty of Interest
- Nihar B. Shah (MLD/CSD): https://www.cs.cmu.edu/~nihars/
- Atoosa Kasirzadeh (Philosophy; courtesy in S3D): https://www.cmu.edu/dietrich/philosophy/people/faculty/atoosa-kasirzadeh.html
-
Trustworthy Forecasting & Infrastructure Readiness
- Stabilizing time-series models (e.g., SARIMA-aware LSTMs), seed-sensitivity studies, and seasonality/heteroskedasticity controls for energy + compute demand planning.
- Reproducibility guarantees (multi-seed protocols, deterministic configs, artifact logging).
-
Evaluation Science & Auditing (ICSD-aligned)
- Methods for robust evaluation, peer-review modeling/assignment, and bias auditing; bridging formal theory with applied pipelines and data-centric checks.
-
Safe RL & Agents (Safe AI Lab-aligned)
- Safety in control/decision-making under uncertainty; sim-to-real considerations; verification hooks in RL pipelines.
-
Community
- This repository aims to pull links together, and organize a format that brings faculty, students, and researchers closer together.
- Thesis will likely be a blend of the above 4.
- Apply Intelligent Systems and Robotics and Probability/Simulation to a relevant project, build upon capstone and deepen knowledge, await for feedback.
Reach back out to Dr. Shah, show progress, and see next steps. Try to align early and begin research.
- Intro to Machine Learning
- Statistics
- Convex Optimization
- Research
- Begin on work proposed in Research.
- Rebuild forecasting pipeline with stricter determinism; publish a reproducibility report (R² distributions vs seeds; calibration).
- Draft an evaluation taxonomy for LLM auditing (annotation robustness; agreement modeling; reviewer assignment framing).
- Probabilistic Graphical Models
- [Insert branch]
- Machine Learning in practice
- [Insert branch]
- Machine Learning with graphs
- [Insert branch]
- Continue into research, build thesis foundations.
- TBD based off lessons learned from previious semester, probably research.
- Safe RL prototype in a simulator (risk constraints, verification hooks).
- Submit a short workshop paper (reproducibility or evaluation).
- To be updated.
- Deep Reinforcement Learning, Graduate Artificial Intelligence.
-
With Prof. Nihar Shah: evaluation science, peer-review modeling, theory with deployments.
-
With Prof. Atoosa Kasirzadeh: AI governance/ethics methods integrated with technical pipelines and evaluation.
-
With Safe AI Lab: verification-minded RL/agents and safety for high-stakes systems.
Maintain running notes, alignment, and annotations.
To be updated
MSDS Alignment Project (baseline + direction): https://github.com/EthanNorton/MSDS-alignment/tree/ai-alignment-problem