My Experience
I'm an experienced quantitative / software development professional. My work provides the "brains" of products for a wide variety of industries. Industries I've served include:
- Defense
- Drug discovery
- Automotive
- Semiconductors
- Quality control and inspection
What's Unique About Me?
In addition to deep and broad technical expertise, I have considerable experience with project management, start-up pitches, presentations to executives, and other critical business functions.
Core Areas of Expertise
Machine Learning
I have designed and continued to extend the core machine learning platforms for various companies. What separates me from many others is that I have extensive experience using ML in production environments. I have considerable "battle tested" experience deploying models in unconstrained scenarios.
Sensors, Signal Processing, and Computer Vision
I have extensive experience working with a wide variety of sensors- RGB cameras, infrared cameras (LWIR, SWIR), UV / SXR cameras, microphones, inclinometers, accelerometers, GPS, FARO laser scanners / trackers, and other sensors. This includes performing measurements, creating autonomous decision making algorithms, and creating sensor calibration processes.
Applied Mathematics and 3D Geometry
I have extensive experience working with 3D measurement products, creating 3D algorithms, and working with optics. For difficult applied mathematics and scientific computing problems, I am typically the "end of the line" technical resource.
Technical Project Management
I have been PM on several multi-million dollar projects involving teams of 10+ scientists, engineers, and other technical professionals. Some of my projects have received innovations awards and other honors. I am proficient running projects with both Agile and waterfall approaches. I am well versed with public speaking, giving presentations to investors, executives, and other key decision makers.
Data Science / Data Analysis
Statistical quality control, statistical significance, anomaly detection, root cause analysis, multivariate analysis, time series analysis- these are all in my wheelhouse. I also have considerable experience creating analysis tools to help your in-house personnel be more productive.
Software Development
Scientific computing software that I wrote / architected is in regular use by thousands of professionals worldwide. I am proficient in a variety of programming languages (Python, MATLAB, C++, C, C#, and others). I am familiar with various source / version control systems, including git, SVN, TFS, and others. I have developed software for a variety of platforms, including: PCs, DSP, ARM, cloud computing (AWS), and others.
MolPred
I was the architect and lead implementer of MolPred. MolPred is a Python platform for applying supervised machine learning to SAR (Structure - Activity Relationship) modeling for drug discovery. It was designed for small molecule drug discovery / lead candidate selection, but it could be adapted for use in a variety of other applications. MolPred was originally developed for scenarios where data is not very plentiful, though it can be used for scenarios where data is quite plentiful.
Take a look at my code here: https://gitlab.com/bdagostino/molpred
My Work History
2012 - present: Algorithms / Machine Learning Consultant, AutonoMee LLC
I work as a consultant for-hire in the fields of machine learning, scientific computing, and software development. I have helped clients large and small to create entirely new product lines and to significantly improve existing products.
2017 - 2019: Director of Machine Learning, Spektron Systems
I lead the machine learning activities in a machine-learning drug discovery start-up. I architected and continue to extend Spektron’s core predictive modeling software platform, which shows superior predictive skill to conventional drug discovery platforms. The platform is an in silico virtual screening environment for computational drug discovery that predicts toxicity, efficacy, and other properties for drug candidates prior to expensive and time consuming synthesis and lab testing.
2011 - 2017: Engineering Manager / Sr. Algorithm Engineer, Snap-on Equipment
I served as the principal algorithms expert for a computer vision based product line, while also working as an engineering manager. Extensive work with 3D geometry, signal and image processing, calibration of measurement systems, and the management of development projects with heavily integrated software and hardware components.
2010 - 2011: Sr. Algorithm Scientist, ChemImage Corp.
I created new computer vision / image processing algorithms for infrared imagers used in explosives detection applications. My work was instrumental in landing several multi-million dollar contracts for real-time threat detection applications.
2007 - 2010: Algorithm Engineer, Textron Systems
I performed data analysis and algorithm development for a defense contractor. Signal / image processing, machine learning, and data management for networks of sensors (acoustic, seismic, and image sensors).
2005 - 2007: Scientist, Energetiq Technology Inc.
Laboratory and analytical scientist for a semiconductor industry startup company. Testing and measurements of prototype lithography equipment for semiconductor fabrication. Extensive use of optics, ultra-high vacuum equipment. Extensive image analysis and laboratory automation algorithm / software development.
2004 - 2005: Scientist, Shiva Technologies Inc.
Laboratory / analytical scientist for a materials testing lab. Helped customers to identify impurities and contamination. Performed GDMS (Glow-Discharge Mass Spectrometry) analysis. Extensive use of UHV equipment. Database organization with MS Access.
Specific Skills
Machine Learning Techniques and Methods
I am well versed in a wide variety of machine learning methods, including but not limited to: deep learning, GAN, CNN, graph convolutional ANNs, SVM, PLS, kNN, decision trees, XGBoost, naive Bayes, logistic regression, and others. I'm a big proponent of bagging and ensemble methods.
Mathematical Techniques
One of my main interests is in applying slick mathematical techniques to help solve real-world problems. Some of the techniques I've used are: gradient descent, Levenberg-Marquardt, bundle adjustment; Nelder-Mead Simplex, adaptive grid searches, PCA, GMM, HMM, and RANSAC, among many others.
Programming Languages, Frameworks, and Tools
I've used a large number of programming languages and can pick up new ones pretty quickly as needed. I'm experienced in: Python, MATLAB, C, C++, C#, R, LabVIEW, and others. I'm experienced in OpenCV, numpy, Keras, scikit-learn, Pandas, and other scientific computing libraries. I have experience with SQL.
Signal / Image Processing Techniques
Apart from standard signal / image processing techniques (convolutions, FFTs, Sobel / Canny edge detection), I've used interesting methods: MUSIC, Kalman filters, adaptive thresholding algorithms, contour tracing and interpolation, target tracking algorithms, INS navigation algorithms; sensor fusion, and others.
Select Publications
Some Patents
USPTO Patent 9,626,559: Target marking for secure logo validation process
This patent is about using computer vision and machine learning techniques to make product counterfeiting significantly more difficult.
USPTO Patent 9,212,907: Short rolling runout compensation for vehicle wheel alignment
This patent is about hardcore numerical linear algebra techniques to achieve extremely accurate measurements in a low-cost angular measurement system. The core of the patent is about measuring and compensating for the effects of runout in a hybrid camera / inclinometer based measurement system, using the absolute minimum quantity of data.
USPTO Patent 10,072,926: Wheel aligner with advanced diagnostics and no-stop positioning
This patent pertains to the design of a breakthrough measurement system for vehicle wheel alignment. The innovation is in using 3D geometry / computer vision techniques to automatically detect and compensate for measurement errors.
USPTO Patent 8,994,934: System and method for eye safe detection of unknown targets
This patent is for a laser-based spectroscopic sensor that uses computer vision / machine learning algorithms to perform pedestrian detection for active surveillance and safety monitoring.
USPTO Patent 9,982,998: Rolling virtual wheel spindle calibration
This patent is about iterative, nonlinear least squares techniques and measurement systems that provide geometrical measurements for vehicle wheel alignment applications.
Some Papers
The vast majority of my work is confidential, but I do have some published research papers.
Development of an infrared imaging classifier for UGS
Abstract: We show design and performance results for an Unattended Ground Sensors (UGS) Automatic Target Recognition (ATR) target classifier using infrared (IR) imagery. Our goal was to develop a basic ATR capability to separate human vs. animal vs. vehicle vs. non-target. Our current UGS video capability accurately detects tracks and transmits targetcentered long wave infrared and visible imagery to a base station. We demonstrate an ATR capability to classify and transmit only targets of interest to the user while excluding others. We describe the ATR development process which includes data collection, building a truthed dataset, feature development, classifier training and performance evaluation.
Proc. SPIE 7693, Unattended Ground, Sea, and Air Sensor Technologies and Applications XII, 76930K (7 May 2010); doi: 10.1117/12.851808;
Abstract: Formulating high sensitivity and high resolution EUV Resists is a critical issue gating the adoption of EUV lithography. The ability of resist manufacturers to quickly screen outgassing rates and sensitivity of EUV resists will facilitate faster formulation of a production-ready EUV photoresist. The high power and low cost per watt of the Energetiq EQ-10 light source enables relatively simple designs without complex optics to deliver relevant data efficiently. Because the source operates without electrodes, a significant source of contamination is removed, further simplifying the design of exposure systems. Data will be presented from two prototype exposure systems.
Formal Education and Training
1998 - 2002: BS, Physics, Syracuse University
Graduated Magna Cum Laude
2007 - 2009: MS, Applied Mathematics, UMASS Lowell
Graduated with Honors
