Steve Krawczyk Email & Phone Number

Petaluma, California, Us

Petaluma, California, Us

General tasks:• Oversee small research team, with multiple direct reports. Planand deliver R&D responsibilities for the company.• Identify relevant features and build a classification models thatpredicts a loan applicant’s performance.• Develop image processing techniques to identify the validity ofidentity documents using data from a mobile device.Specifics:Complete pipeline developed using Pandas, Numpy, and scikit-learn for both explainable and non-explainable ML models. Deployed models on AWS leveraging Flask.Object segmentation, detection, and recognition models developed for driver's licence images using Tensorflow and Keras. Significant improvement over previous methods leverage deep networks with transfer learning. Additionally, GAN models were developed for one class authentication.

Ann Arbor, Mi, Us

General tasks:• Develop proof of concept to predict user identity from datastreams from workout equipment.• Time series data analysis to predict repetitions, recommendedweight setting, etc.Specifics:Implemented kalman filtering on noisy sensor data (accelerometer and load sensors). This improved accuracy of rep counting.Developed user authentication pipeline with for two separate methods: a sliding window BOW model and a LSTM model using DCT features.

Ann Arbor, Michigan, Us

General tasks:• Design and implement algorithms related to expert systems,cognitive architectures, machine learning and machine vision.• Build algorithms designed to maximize situational awarenessand estimate controller workload for autonomous vehicles.• Integrate a verb learning and vision system with a groundrobot, learning spacial relationships among specific targets.Specifics:Leveraged SOAR to add additional actions for a robotic arm interacting with objects.Assisted in translating SOAR C components to Java.Developed Java-based demo that assisted in controlling multiple autonomous vehicles to improve situational awareness.

San Francisco, Ca, Us

General tasks:• Apply machine learning algorithms for directed advertising at very large scale using map reduce.• Optimize which creative to display for a bid request by analyzing features of past and the current bid request.• Cluster converters of a customer into sub-populations where each population can be targeted appropriately.Specifics:Developed map reduce jobs with leveraging both NB and SVM models leveraging web activity of users to predict conversion for various customers. These models typically consisted of millions of records of very sparse data.Leveraged LDA to cluster a client's customers. The resulting clusters were leveraged to refine the creatives displayed for each group, improving conversion performance.

Petaluma, California, Us

General Tasks:Research and develop algorithms for face recognition with a focus on DVR imagery.Work with identity documents to research methods for improved face recognition on such imagery and basic methods of document authentication.Specifics:Built a FR algorithms that specifically handled comparing driver license photos. This algorithm outperformed the SOA on this image of this type.Built various authentication methods for drivers licenses photos leveraging hough transforms, SURF features in a BOW model, FFT, and DTW.

Dresden, Saxony, De

General tasks:Researched techniques to increase the signal to noise ratio of images using video sequences.Explored methods regarding image registration and fusion.Specifics:Leveraged Kalman filtering to improve eye tracking performance in video sequences.Designed an algorithm to build the best candidate face for matching from a video sequence.

General tasks:Worked under the supervision of Dr. Behrooz Kamgar-Parsi.Research the development of a similarity measure that mimics how humans classify faces.Specifics:Develop image processing algorithms for lighting and shadow normalization.Investigated a neural network classifier and several methods of seeding training data.Implemented infrared face recognition and its fusion with the visible domain.

Doctor Of Philosophy - Phd, Computer Science

Master'S Degree, Computer Science

Bachelor'S Degree, Computer Science