Who I am, and how I work
What I build, and what I'm exploring next.
I'm Rustam (Russ) Mukhtarov, an ML/AI Engineer and M.S. candidate in AI/Data Science at San Francisco State University, with 3+ years building and deploying production ML pipelines. I turn models into usable software: APIs, evaluation pipelines, Dockerized services, and clean interfaces.
My recent work spans citation-grounded RAG systems, deployed ML inference APIs, and 3D deep learning for medical imaging. I care about honest evaluation, reproducibility, and shipping systems people can actually run — tests, CI, and Docker included.
What I do
I take ML from notebook to deployed service: data prep, training, evaluation, serving, and the interface on top. I'm comfortable moving across model pipelines, backends, and frontends — prioritizing correctness, reproducibility, and measurable outcomes.
Current focus
- Hybrid retrieval (BM25 + dense) and evaluation harnesses for RAG systems.
- 3D U-Net segmentation of micro-CT data for my graduate thesis.
- Making deployed models trustworthy: grounding, abstention, and explainability.
Graduate thesis
AI-Assisted Medial-Axis Modeling of the Human Fallopian Tube — building a 3D deep-learning pipeline (PyTorch, MONAI) for lumen segmentation (target Dice ≥ 0.85) and centerline extraction with geometric descriptors (radius, curvature, torsion). Includes a Python toolkit (VTK/VMTK/SciPy) for quantitative analysis and dataset reproducibility. Advised by Prof. Kazunori Okada.
Principles I work by
Beyond work
I lift, swim, and play basketball to reset my brain. Music (piano & guitar) keeps me curious and patient. I enjoy learning languages and exploring SF's food scene.