AI Safety Researcher · Data Scientist · Physics PhD
Kian Maleki, PhD
I do empirical AI-safety research on reasoning models and build machine-learning and document-intelligence systems, with a research scientist's habit of asking: What can fail, how do we know, and how do we measure it? My work connects physics, careful evaluation, and practical AI tools.
Focus
Three areas, one thread: careful modeling.
My background moves from physics and numerical simulation into applied machine learning and, most recently, empirical AI safety. The connecting idea is the same: build models, test their limits, and explain when they should be trusted.
-
AI Safety & Reasoning Models
- Chain-of-thought faithfulness
- answer commitment
- pre-registered causal experiments on reasoning traces
- prefix interventions
- resampling studies
- evidence about what visible reasoning actually reveals
-
Machine Learning & Model Evaluation
- Classifier ambiguity
- calibration
- threshold-dependent behavior
- finite-horizon learning-curve prediction
- conservative extrapolation
- experiment tracking
- reproducible benchmarks
-
AI Harnesses & Automation
- Agent harnesses and orchestration
- tool use with verification loops
- automating research workflows
- industry process automation
- evaluation and monitoring
- reproducible pipelines
Selected work
Projects with a research spine.
These projects are designed to be more than portfolio demos. They test reliability, expose failure modes, and connect mathematical ideas to usable AI systems.
-
View project ↗
Reasoning Trace Experiments
AI SafetyA study of where a reasoning model's answer takes shape, resampling its chain of thought at block boundaries.
-
View project ↗
Decided Mid-Thought
AI SafetyA pre-registered causal study of chain-of-thought faithfulness in a reasoning model: prefix interventions and 500 resampled continuations test whether visible reasoning reveals when the final answer becomes determined.
-
View project ↗
Ambiguity Range Framework
Model EvaluationA diagnostic toolkit for evaluating binary classifiers beyond a single threshold, with ambiguity-focused metrics, calibration analysis, confidence intervals, and statistical testing.
-
View project ↗
Sequence Acceleration Benchmark
Learning CurvesA reproducible benchmark for finite-horizon gradient-boosting learning-curve prediction across many convergence regimes, noise levels, and extrapolation methods.
-
View project ↗
PharmaDoc-AI
Document AIA pharmaceutical document-intelligence and RAG project that combines PDF extraction, structured information retrieval, document question answering, and evaluation workflows.
-
View project ↗
LumberLex
Data NormalizationA lumber product-name normalization engine that maps inconsistent seller strings to canonical structured outputs using fuzzy matching, dimension extraction, treatment detection, confidence scoring, a Streamlit interface, and chatbot-style explanations.
-
View project ↗
Heart Disease Prediction Pipeline
MLOpsAn end-to-end MLOps pipeline with DVC, MLflow, pytest, GitHub Actions, model-performance checks, and drift monitoring with Evidently.
Research
Physics roots, machine-learning branches.
My research background spans quantum materials, crystal-field theory, noncollinear magnons, molecular and defect-mediated transport, dark matter phenomenology, modern ML diagnostics, and, most recently, empirical AI safety on reasoning models.
-
[1]
Nano Letters, 2025. A General and Modular Approach to Solid-State Integration of Zero-Dimensional Quantum Systems. Nano Lett. 25, 13787–13794. doi:10.1021/acs.nanolett.5c03125
-
[2]
Physical Review B, 2025. Crystal fields, exchange and dipolar interactions, and noncollinear magnons of erbium oxide. Phys. Rev. B 112, 064434. doi:10.1103/v6n3-srt5
Research interests
- AI safety
- Reasoning models
- CoT faithfulness
- Classifier ambiguity
- Learning curves
- Calibration
- Scientific ML
- Document intelligence
- RAG
- High-dimensional modeling
- Numerical simulation
Experience
A bridge from research to applied AI.
My path combines academic research, teaching, scientific software, and project-based machine-learning training.
- 2025 – present
Independent Researcher
AI automation and LLMsSelf-directed, open-source research and development in AI automation and large language models, with public code and write-ups on GitHub.
- 2026
Technical AI Safety Project
BlueDot ImpactCompleted the program research project, covering Decided Mid-Thought and Reasoning Trace Experiments.
- 2025 – 2026
Resident Assistant Professor of Physics
Creighton UniversityTaught undergraduate physics, developed optics and electronics lab materials, supervised teaching assistants, and integrated AI tools into scientific workflows.
- 2025 – 2026
AI & Machine Learning Training
TripleTenCompleted project-based training across supervised learning, unsupervised learning, deep learning, NLP, computer vision, SQL, Git, and end-to-end ML workflows.
- 2026
AI-Powered Document Intelligence Externship
Pfizer / ExternBuilt retrieval-augmented and document-intelligence workflows for healthcare documents using native PDF extraction, LLMs, and RAG architectures.
- 2018 – 2025
Graduate Research Assistant
University of IowaDeveloped high-dimensional numerical models, optimized scientific simulations, analyzed noisy data, and contributed to peer-reviewed work in condensed matter physics and quantum materials.
- 2015 – 2018
Graduate Research Assistant
Creighton UniversityBuilt simulation-based models using Python and Fortran, including N-body simulation, molecular dynamics, and parallelized scientific computing workflows.
Tools
Technical stack.
- Machine Learning & LLMs
- Scikit-Learn, XGBoost, PyTorch, Transformers, LLM APIs, reasoning-trace analysis, calibration, model evaluation, NLP, RAG, MLflow.
- Engineering
- Python, SQL, FastAPI, Pydantic, Docker, Linux/Bash, Git, GitHub Actions, Jenkins, CI/CD, pytest.
- Scientific Computing
- NumPy, Pandas, Polars, Numba, Fortran, HPC, MPI, numerical optimization, Monte Carlo simulation.
- Document Intelligence
- PDF parsing, information extraction, OCR-aware workflows, structured extraction, evaluation of extraction quality.
Contact
Let's connect.