McAllister P. Blair

B.S. in Data Science, ML & AI Track. Major GPA 3.83, cumulative GPA 3.4. La Jolla, CA.

Coursework: Autonomous Vehicles, Machine Learning, Artificial Intelligence, Data Structures and Algorithms, Linear Algebra, Multivariable Calculus, Statistics, Database Management Systems.

Autonomous lab robotics, DOE Genesis Mission proposal (Sep 2026 to present)

• Own the robotics layer of ArcellAI's instrument-neutral autonomous-lab proposal: a simulated skill that transfers a sample carrier from a rack into a characterization station and verifies correct placement before a measurement runs.
• Specifying the workcell (coordinate frames, planning scene, collision-aware motion with ROS 2 and MoveIt Task Constructor) and tracing successful and failed transfer episodes against a 24-case fault catalog.
• Evaluating learned 6D pose estimation (FoundationPose) against a fiducial-marker baseline for visual placement verification.

Biosignal data quality (June 2026 to Sep 2026)

• Built an end-to-end ingestion, cleaning, and validation pipeline (Python, Pydantic, MNE) for heterogeneous clinical time-series recordings, normalizing inconsistent sources into one validated schema with hash-based provenance.
• Designed a configurable quality engine running automated integrity checks (shape mismatch, missing and duplicate channels, corrupt records) that emits typed, severity-tiered quality events with supporting evidence.

• Designed and deployed a camera-based imitation-learning (behavior cloning) autonomy stack for a 1:10-scale vehicle, training a CNN to map onboard camera frames to steering and throttle.
• Deployed trained models with TensorFlow Lite for real-time inference on a Raspberry Pi 5 running a 20 Hz control loop, improving validation loss 3.4× across iterations.

• Built a 6-DoF follower arm from hardware assembly through deployment, with a custom PS4 controller teleoperation interface for collecting demonstrations.
• Trained an Action Chunking Transformer policy on 50+ demonstrations (PyTorch, Hugging Face, W&B), cutting training loss from 0.35 to 0.06, and validated autonomous pick-and-place on hardware.

• Built a ROS 2 (Jazzy) safety node in C++ that computes per-beam time to collision from 2D LiDAR and odometry and overrides drive commands on the F1Tenth interface at sensor rate.
• Validated against a synthetic LaserScan and odometry fixture covering on-axis, off-axis, and stationary cases, with ground truth derived independently of the node's formulation.

• Designed and built a custom analog front-end (AD620 instrumentation amplifier, discrete active filtering) to capture ECG from surface electrodes.
• Built a real-time acquisition pipeline (Arduino Nano Every to PySerial) sampling at about 190 Hz, with MNE band-pass filtering to remove baseline wander and mains noise.

• Jeep Bot platform team (ROS 2, LiDAR and stereo camera sensing) for autonomous campus navigation, working on obstacle avoidance: local planning, costmaps, Nav2 integration, and a safety layer.