Ben Semerjian, Developer in Portland, OR, United States
Ben is available for hire
Hire Ben

Ben Semerjian

Verified Expert  in Engineering

Scientific Computing Developer

Location
Portland, OR, United States
Toptal Member Since
September 29, 2016

Ben is a seasoned expert in computational science with an emphasis on computer vision. For the past ten years, his passion was to begin hard science quests, pose them as optimization problems, and then solve them with computing machinery. His main strengths are all manners of numeric optimization, C++11 fluency (tasteful, perfectionist, consistent, and poetic source code), and first-rate problem formulation/raw analysis skills.

Portfolio

Self-Employment
Hardware, Scientific Computing, C++, Python, Artificial Intelligence (AI)
HoneyComb Corporation
C++, Optimization, Computer Vision
OGSystems
Optimization, C++, Computer Vision

Experience

Availability

Part-time

Preferred Environment

Git, MSVS

The most amazing...

...thing I did was to write a potent new method method for extracting 3D surfaces from images.

Work Experience

Engineer | Scientist

2015 - PRESENT
Self-Employment
  • Designed and built a low-cost, high-resolution smart imaging system for a high altitude balloon.
  • Built a subpixel accurate video rectification application for an ion propulsion experiment.
  • Developed a vignetting correction and calibration back-end for a camera manufacturer.
  • Designed, built, and engineered a high precision fluid mechanics experiment for a university at one-tenth of the price quoted by established machine shops.
  • Conducted ongoing research in noise reduction, demosaicing, deconvolution, and compressed sensing.
Technologies: Hardware, Scientific Computing, C++, Python, Artificial Intelligence (AI)

Computer Vision Scientist

2015 - 2016
HoneyComb Corporation
  • Built an extremely fast and highly reliable image/map processing pipeline.
  • Invented methods for avoiding UAV-specific imaging problems.
  • Invented a spectral calibration method which someone else happened to publish recently.
Technologies: C++, Optimization, Computer Vision

Computer Vision Scientist

2009 - 2015
OGSystems
  • Built a cluster process using ZeroMQ.
  • Wrote a many different custom optimizations.
  • Found a practical solution to the three view problem.
  • Invented and published a new multi-view stereo method.
Technologies: Optimization, C++, Computer Vision

Pinhole Lens Simulation

This project is related to what we know as pinhole lenses: You can punch a hole in aluminum foil, put it on a camera in front of a sensor (no lens), and it will take pictures. Blurry and with other issues, but it will work. The first cameras used pinhole lenses.

The program I created simulates a description of the blur of a pinhole lens by solving the electromagnetic wave equation. It makes very few assumptions, most notably that you're photographing something "far away" and that the foil is thin. It simulates four holes next to each other arranged in a square.

When running the program, you can copy its output to a spreadsheet and then make a graph. You will see two big spikes and a lot of wiggles. The four pinholes generate four spikes, but you can't see them because they're somewhere else in 3D. The way I solved this is unique, with its advantages and disadvantages, compared to the traditional Fourier approach.

Back-End Map Processing Pipeline

I built HoneyComb's back-end map processing pipeline on my own. The application includes state-of-the-art feature matching, bundle adjustment, compositing, calibration, and GIS output. It outperforms Pix4D for this specific purpose by a significant margin.

A New Variational Framework for Multi-View Surface Reconstruction

https://pdfs.semanticscholar.org/3149/f2f9e3b5a5bdd8e8924fd34c4514892ef14e.pdf
This is a publication that I wrote while working at OGSystems which is the culmination of three years of work on surface reconstruction. The conference that I presented this at has a five percent acceptance rate for surface reconstruction topics.

PeARL™ Flash 3D Processing Pipeline

For over six years, I wrote almost all of the algorithms and pipelines behind OGSystems' 3D processing product. It is currently being gainfully exploited and maintained by others.

StockPileReports.com Back-End

https://www.stockpilereports.com/
I contributed to the launch of StockPileReports.com in 2012 by writing the SfM pipeline and a tool to measure the volume of stockpiles from noisy point clouds.
2007 - 2009

Master's Degree in Mechanical Engineering (Fluid Mechanics)

Portland State University - Portland, OR, USA

2002 - 2007

Bachelor's Degree in Mechanical Engineering

Portland State University - Portland, OR, USA

Libraries/APIs

ZeroMQ, OpenCV, OpenMP

Tools

Git, Adobe Photoshop

Languages

C, C++, Python, C#

Frameworks

Boost, OpenCL

Platforms

Windows, Linux, NVIDIA CUDA

Paradigms

Unit Testing, Agile Software Development

Other

Algorithms, Back-end Development, Computer Vision, Numerical Methods, Scientific Computing, Computer Science, Artificial Intelligence (AI), MSVS, Optimization, Hardware, Computer Graphics, Simulations

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring