Mustafa Shaikh

I work with deep learning, computer vision, and robotics.

I'm currently working on a multi-object tracking problem for animal and cell video data using deep learning, at a Computational Neuroscience Lab at Salk Institute for Biological Studies. The project, DREEM, is fully open source.

Previously, I worked in industry for ~5 years in large organizations in a data science role. I worked closely with business stakeholders and led projects end to end.

Fun fact: I didn't always work in the world of data and machine learning. My undergrad specialization was in energy systems, so in 2015, I spent a year working as a field engineer helping to build a 300 MW combined cycle power plant. If you're interested in learning more about how our electricity is generated, ask me about this!

I've also been exploring LLMs, and have created a ready-to-use LLM workflow that is integrated with Cursor. Check it out here.

Publications

Projects

An LLM workflow with a finetuned Llama 3.2 1B Instruct that retrieves relevant research papers from Arxiv based on user chat queries. Integrated with Cursor via MCP.

Landmark and trajectory estimation from calibrated stereo camera features and IMU measurements.

Particle filter for occupancy and texture mapping using LiDAR, RGBD camera, and IMU data.

Classical 3D motion planning with obstacles.

A barebones String library with a buffer manager.

A general purpose archive tool with compression, add, remove, and retrieve functionality.

An in-memory JSON model with a query language to select, filter, and aggregate data.