Skip to content
View Rustam751's full-sized avatar

Block or report Rustam751

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Rustam751/README.md

Typing SVG

Rustam (Russ) Mukhtarov

Applied ML / Computer Vision / AI Product Engineer M.S. Artificial Intelligence & Data Science @ San Francisco State University San Francisco, CA

I build applied AI systems across medical imaging, computer vision, UAV robotics, and human-AI product prototypes. My work combines model development, data preprocessing, evaluation, and product-focused engineering.


Featured Projects

RAG Document Assistant

Production-style retrieval-augmented generation system for medical, legal, and academic PDFs. Every answer is grounded in cited sources (file + page + verbatim quote), and the model abstains instead of guessing when documents don't contain the answer.

  • Stack: Python, FastAPI, ChromaDB, Anthropic Claude API, Streamlit, Docker, GitHub Actions
  • Focus: RAG, grounded generation, hallucination guards, LLM evaluation, API engineering
  • Results: 0.87 retrieval hit rate, 100% abstention on unanswerable questions, and zero fabricated facts on a 20-question benchmark — with documented failure analysis of the two retrieval misses
  • Solo project: designed, built, tested (29 offline tests + CI), evaluated, and deployed end-to-end

Repository: https://github.com/Rustam751/rag-document-assistant


Medical Image Disease Classifier API

Production-style medical image classification service that extends my Alzheimer's PET research into a deployable ML system. It accepts a scan image over HTTP and returns a predicted class with a full probability breakdown, input validation, clear error handling, and a Grad-CAM explanation of where the model looked.

  • Stack: Python, FastAPI, PyTorch, Streamlit, Docker, GitHub Actions
  • Focus: ML engineering, model serving, reproducibility, input validation, explainability (Grad-CAM), preventing train/serve skew
  • Results: REST API + Streamlit demo, 20 tests with CI, Dockerized runtime, and a ~0.4 MB self-describing checkpoint that serves on clone. Reaches 94.1% accuracy / 0.94 macro-F1 on a reproducible 3-class benchmark — trained on a clearly-labeled synthetic dataset (non-clinical) so results reproduce without restricted data, with real datasets plugging in via the same ImageFolder interface
  • Solo project: designed, built, tested, evaluated, containerized, and documented end-to-end

Repository: https://github.com/Rustam751/medical-image-classifier


Alzheimer's PET Classification

Deep learning pipeline for classifying Alzheimer's disease stages from Amyloid PET images using transfer learning.

  • Models: VGG16, InceptionV3, ResNet50
  • Focus: Medical imaging, CNNs, transfer learning, preprocessing, evaluation
  • Results: VGG16 and InceptionV3 achieved above 95% test accuracy in the project setup
  • My contribution: Model experimentation, preprocessing workflow, evaluation, and result analysis

Repository: https://github.com/Abdoul1996/Alzheimer_Classification


DroneHunter / UAV Detection

Computer vision and tracking project connected to UAV competition work, involving detection, tracking, and real-time robotic vision.

  • Stack: YOLO, OpenCV, object tracking, UAV systems, simulation/field testing
  • Focus: Real-time computer vision, drone detection, robotics integration
  • My contribution: UAV detection workflow, testing, model/inference support, and competition implementation

Repository: https://github.com/keremtatlici/dronehunter_teknofest2023


DriveShare — Parking Marketplace Prototype

Peer-to-peer driveway rental platform where homeowners can list unused driveway space and drivers can reserve affordable parking.

  • Stack: Next.js, React, Tailwind, PostgreSQL, UX research
  • Focus: Product engineering, full-stack prototype, HCI, user flows
  • My contribution: Product concept, UI/UX flows, web prototype, research pages, and driver/homeowner experience design

Repository: https://github.com/Rustam751/driveshare-app Live Demo: https://driveshare-app.vercel.app/


Tennis Court Line Detection

Image-processing pipeline for detecting tennis court lines from match images.

  • Stack: Python, OpenCV, image processing
  • Focus: Line detection, preprocessing, evaluation, computer vision fundamentals

Repository: https://github.com/mucahitkalebasi/tennis-line-detection-img-processing


Technical Skills

  • Machine Learning / AI: PyTorch, TensorFlow/Keras, Scikit-learn, CNNs, Transfer Learning, MONAI, YOLO, OpenCV
  • GenAI / LLMs: LangChain, LlamaIndex, Hugging Face, OpenAI API, RAG, FAISS, Chroma, Prompt Engineering
  • Engineering: Python, TypeScript, JavaScript, C++, SQL, Flask, React, Next.js, Node.js, PostgreSQL
  • Tools: Docker, Git, Linux, Bash, MATLAB, Raspberry Pi, Gazebo, DroneKit
  • Data Science: pandas, NumPy, SciPy, Matplotlib, statistical modeling, data preprocessing, evaluation metrics

Experience Highlights

  • AI Team Lead — RevSolz Corp. Led data science interns and built ML pipelines for oil-industry operational data.
  • Graduate Thesis — SFSU Working on AI-assisted medial-axis modeling and 3D deep learning for medical imaging.
  • Teaching Assistant — Theory of Computing, SFSU Supported 30+ students with Python setup, lab support, grading, and course materials.

Contact

Pinned Loading

  1. Abdoul1996/Alzheimer_Classification Abdoul1996/Alzheimer_Classification Public

    Jupyter Notebook

  2. mucahitkalebasi/tennis-line-detection-img-processing mucahitkalebasi/tennis-line-detection-img-processing Public

    Python 1

  3. keremtatlici/dronehunter_teknofest2023 keremtatlici/dronehunter_teknofest2023 Public

    Python

  4. driveshare-app driveshare-app Public

    DriveShare — a peer-to-peer parking marketplace prototype where homeowners list unused driveway space and drivers book affordable, nearby parking. Full HCI-driven product flow with driver/homeowner…

    TypeScript

  5. rag-document-assistant rag-document-assistant Public

    Grounded RAG assistant for PDFs — FastAPI, ChromaDB, Claude with cited, hallucination-guarded answers

    Python

  6. medical-image-classifier medical-image-classifier Public

    Production-style medical image classification API — FastAPI + PyTorch model serving with preprocessing, Docker, tests, CI, and Grad-CAM explainability. Educational/portfolio project inspired by Alz…

    Python