CV — Machine Learning Engineer

Download ML Engineer PDF  |  All resumes  |  Robotics SWE view

PDF is auto-generated from this page by CI on every content change.

TL;DR — ML engineer with applied research depth in computer vision, 3D point-cloud detection, and multimodal sensor fusion. Trained and deployed CNN models on AWS for NYC infrastructure inspection and traffic monitoring; comfortable owning the full loop from raw multi-sensor data → labeled datasets → model → cloud / edge deployment.

Core skills

Models & methods 2D & 3D CNNs (object detection, segmentation), contrastive learning, multimodal fusion (LiDAR + RGB + thermal / hyperspectral), background subtraction and tracking, classical computer vision (OpenCV), control-theoretic models (PID, fractional-order) for closed-loop CV systems.

ML platforms & infrastructure PyTorch, TensorFlow, CUDA, AWS (S3 for data, EC2/SageMaker for training, model hosting), Docker, NVIDIA Jetson (Nano / TX2 / Xavier) for edge inference, WebODM-based deployment platform for end-user delivery.

Data & pipelines End-to-end ingestion from real-world sensors (fixed LiDAR + camera, drone payloads, robotic platforms), dataset versioning, format conversion across sensor configurations, ROS / MCAP / Protobuf, labeling and quality workflows, training-on-A → inference-on-B portability.

Languages Python (primary), C++, MATLAB, Fortran.

Selected ML projects

CNN-Based 3D Detection on Highway LiDAR · 2024–present

AI & Mobility Research Lab, CCNY · NYC DOT collaboration

• Evaluating and tuning CNN-based 3D object detectors on fixed-LiDAR highway data covering vehicles, motorcycles, bicycles, and pedestrians.
• Built a configuration-portable pipeline: train on one dataset, infer on another with different sensor parameters — handling beam count, mounting geometry, and timing differences.
• Multi-frame point cloud reconstruction to densify per-vehicle representations for downstream classification.
• Co-authored MobiSPC 2025 papers on (a) how many LiDAR beams are enough for vulnerable road user detection, and (b) a survey of sensing perspectives for VRU monitoring.

Bridge Defect Detection & Cloud Deployment · 2022–present

CCNY Robotics Lab

• Trained CNNs for crack / spalling / stain detection on robot-collected imagery; pushed models to AWS for scalable inspection.
• Built a custom WebODM-based platform integrating automated segmentation, 3D reconstruction, interactive visualization, and crack measurement.
• Selected for IEEE IROS 2025 presentation (Feng, Shang, et al., IEEE T-ASE, 2025).

Contrastive Learning for Robust Defect Mapping · 2024

• Contrastive learning approach for impact-echo defect mapping on concrete slabs, robust across acquisition conditions. (Hoxha, Feng, …, Shang, et al., 2024.)

Vision-Based Robotic Control · 2015–2022

• Drone visual servoing with fractional-order controllers — closed-loop computer vision at the edge (Raspberry Pi / embedded Linux), demonstrating control under longer-than-typical sampling periods.
• Girder detection on NVIDIA Jetson as part of the BIRDS bridge-inspection drone.
• SmartCaveDrone — sense-and-avoid + GPS-denied navigation using onboard vision in unstructured environments.

Earlier signal-processing and CV work · 2010–2014

• Visual reconnaissance path planning algorithms; object tracking quadrotor on ROS + AR.Drone; data/image transmission systems (patented).

Education

• PhD, Civil Engineering (Transportation) — CCNY, NY, USA · 2025–present
• PhD, Pattern Recognition and Intelligent Systems — Northeastern University, China · 2013–2020
• Exchange PhD — UC Merced · 2015–2017
• MEng, Pattern Recognition and Intelligent Systems — Northeastern University, China · 2011–2013
• BEng, Automation — Northeastern University, China · 2007–2011

Experience

• 2025–now — Postdoctoral Scholar, AI & Mobility Research Lab, CUNY City College
• Dec 2022–Dec 2024 — Postdoctoral Researcher, CCNY Robotics Lab
• Jan 2020–Nov 2022 — Postdoctoral Fellow, Missouri University of Science and Technology
• Jun–Aug 2024 — Instructor, Computer Engineering Summer Academy (AI module), Vaughn College
• 2023–2024 — Adjunct, Vaughn College (Principles of AI, Robotics Mechanics & Control)
• 2015–2017 — Lecturer / Junior Specialist, UC Merced

Teaching (ML / AI relevant)

• Principles of AI (Vaughn College, SBC 012)
• Principles of Research — AI (Vaughn College, SBC 014A)
• Robotics, Mechanics and Control (Vaughn College, MCE 355)

Selected publications

1. Bo Shang et al. Sensing Perspectives on Vulnerable Road User Monitoring for Traffic Safety: A Survey. MobiSPC 2025.
2. Bo Shang et al. How Many Beams of LiDAR is Enough for Detecting Vulnerable Road Users? MobiSPC 2025.
3. Jinglun Feng, Bo Shang et al. Robotic Inspection and Data Analytics to Localize and Visualize the Structural Defects of Concrete Infrastructure. IEEE T-ASE 2025. Selected for IROS 2025.
4. Hoxha, Feng, …, Shang, et al. Contrastive Learning for Robust Defect Mapping in Concrete Slabs using Impact Echo. 2024.
5. Zhang et al. Code-specified early delamination detection and quantification in a RC bridge deck. 2025.

Full list on the Publications page and Google Scholar.

Reviewer service (ML / robotics venues)

IEEE T-CST, ICRA, IROS, ICUAS, IEEE MFI, J. of Intelligent & Robotic Systems, Mechatronics, Control Engineering Practice, Nonlinear Dynamics, ISA Transactions, IET Control Theory & Applications, IJARS.

Why this role fits

I bring an applied ML profile shaped by real sensor hardware: I’ve trained the model, debugged the calibration that made the dataset trustworthy, deployed it to AWS or to a Jetson on a robot, and shipped the result to a city agency. If you want an ML engineer who is comfortable upstream and downstream of model.fit(), that’s me.

Looking for the Robotics SWE view, the full academic CV, or all resumes?