About
I'm a Computer Science student at Carnegie Mellon University in Qatar interested in machine learning, 3D computer vision, robotics, and practical software systems. My research has focused on benchmarking metric depth estimation for construction perception and developing LLM-based tools for notebook assessment.
Skills
Research
- Metric Depth & 3D Mapping Undergraduate summer research with Prof. Gianni Di Caro at CMU-Q. Built a synthetic industrial rebar dataset with exact ground truth and a metric-depth benchmarking pipeline. Compared monocular depth models with stereo and multi-view approaches and analyzed failure modes on repetitive industrial geometry.
- LLM Autograding for Notebook Assessments Built a rubric-guided LLM grading pipeline for flow-structured Jupyter notebook assignments, with path-aware assessment, schema-constrained outputs, Gradescope export, and review tooling. Developed the full autograder as a Course Assistant, used it in a CMU-Q course, and extended the work into a research paper.
Publications
- Benchmarking Metric Depth for Construction Perception: Industrial Rebar Scenes with Exact Synthetic Ground Truth First author · IROS 2026 Construction Robots Workshop. Presented and selected as a spotlight among the top three accepted submissions.
- Beyond Unit Tests: Path-Aware Rubric-Guided LLM Autograding of Notebooks First author · Accepted at IEEE TALE 2026; not yet published.
Projects
- Autonomous Robotics Independent Study ROS and Gazebo work from an autonomous robotics independent study: control, localization, mapping, and planning (PID, bug navigation, VFH, path following).
- City Image Classifier Trained a CNN city image classifier with TensorFlow/Keras, achieving over 80% accuracy. Built a Selenium pipeline to collect and assemble the image dataset.
- ML Notebook Autograder Path-aware, rubric-guided LLM autograder used in a CMU-Q course. Parses notebook submissions and exports grading results to Gradescope and Excel, with a CLI and FastAPI web UI.
- Smart Dispatch Simulation Built a multi-agent dispatch simulation at the Snoonu Hackathon using combinatorial bidding. Reduced the required fleet size by 14% while maintaining delivery times under 15 minutes in the simulation.
- Graph Theory Library Python library for graph structures and algorithms.
- Athan GNOME Extension GNOME Shell extension fork for Athan prayer times.
More on GitHub, including LaTeX course notes.
CV
View or download my resume below.
On mobile, use the buttons above to open or download the PDF.
Contact
- Email salmanhajizade2005@gmail.com
- University email shajizad@andrew.cmu.edu
- GitHub github.com/sal0-h
- LinkedIn linkedin.com/in/salman-hajizada