Collaborating with one of the world's largest automotive companies to develop a transformer-based end-to-end self-driving car for use in specific driving situations.

__Perception__

• Employed multi-modality multi-sensor multi-view learning-based fusion frameworks.

• Generated bird's-eye-view (BEV) feature maps using both projection (3D->2D, sampling, pulling) and unprojection (2D->3D, splatting, gaussian splatting, pushing) methodologies.

• Incorporated temporal information into BEV feature maps.

• Estimated depth using indirect supervision and direct lidar supervision.

• Utilized intrinsic, extrinsic, and distortion parameters from fisheye camera calibration for BEV generation.

__Detection__

• 2D and 3D object detection of static and dynamic objects.

• Tracked historical trajectories of agents such as vehicles and pedestrians.

• Predicted future trajectories of agents.

• 2D and 3D instance and semantic segmentation.

• Detections of directed and undirected vectorized map components such as lanes.

__Planning__

• Developed planning modules that take into account upstream features to predict paths for the ego vehicle.

__Evaluation__

• Evaluated predicted paths using both popular metrics (ADE, Collision Rate) and more in-depth customized proprietary metrics. Some of our models achieved performance exceeding public state-of-the-art models on key motion planning metrics.

• Evaluated object detections using both popular metrics (L2, IoU) and more in-depth customized proprietary metrics.

Programming language used: Python.

Commonly used frameworks and libraries: pytorch, numpy, mmdetection, shapely, pyquaternion, matplotlib, opencv, pydantic, hydra, and more.

Development environments used: Docker, Virtual Machines, AWS, Amazon EC2 UltraClusters.

Team roles taken on have included: researcher, engineer/developer, interpreter/translator, client communication liaison, and project manager.

Researched and developed custom state-of-the-art AI solutions to automate business processes at client companies. Trained other engineers on how to do the same. Also single-handedly created a comprehensive computer vision framework used by all internal engineers that accelerated proof-of-concept development time from months to days.

I lead over forty unique projects in industries including farming, manufacturing, construction, medical, utility, transportation, forestry, warehouse, and broadcasting.

A few examples (full list available on request):

• Asbestos detection.

• Checking the quality of harvested crops (size, shape, texture, color).

• Checking for abnormalities in products on manufacturing lines.

• Checking if construction workers are wearing their safety harnesses.

• Measuring the diameters of cylindrical objects in images.

• Classifying pills.

Data types used were mostly visual and included images, videos, and depth maps (stereo and LiDAR). Dataset sizes ranged from nothing to about 10TBs.

Deployment environments used include the cloud (Azure VMs, containers, functions, etc), an iPhone, an Android-based augmented reality headset, and Jetson embedded devices (Nano 2GB, Nano 4GB, Xavier NX).

Programming languages used were mostly Python, but also a little bit of HTML, Javascript, CSS, Swift, and Kotlin.

Libraries used include NumPy, OpenCV, Scikit-Learn, Pytorch, Pandas, and Matplotlib

HECM Senior Home Financing is a multi-state mortgage brokerage based in Orlando, Florida. NMLS #1415881.

My tasks included but were not limited to:

• Weekly payroll for up to about 20 employees

• 401k management

• Reporting and filing of state taxes (annual reports, sales taxes, franchise taxes, unemployment taxes, etc)

• Reporting and filing of federal taxes (1120S, K1s, etc)

• Business registration and maintenance in multiple states

• Accounting/Bookkeeping (A/R, A/P, depreciation, reconciliation, monthly reports, etc)

• Financial analysis and budgeting (cost accounting, capital budgeting, etc)

• Working with state financial regulators in each state to maintain compliance

• HR compliance

• Management of loan officer's professional licenses through NMLS

• Tech support

• Creation and maintenance of multiple websites (HTML, CSS, Javascript, Wordpress)

• IT systems administration (setting up, maintaining, and integrating IT systems such as Office 365, CRMs, and our network of websites)

Taught English to children from the ages of three to fifteen. Also taught robotics (using Lego Mindstorms) and programming fundamentals to kids from the ages of three to fifteen.

Taught English, Math, Drama, and Science to kids aged four to ten.