AREA OF EXPERTISE:

Computer Vision, Deep Learning, Machine Learning

SUMMARY:

• 10+ years of experience covering research, development, deployment and product support of computer vision application in real world.
• Experience with large scale data analysis, machine learning model training and performance validation.
• Familiar with the recent advancement in computer vision and deep learning research, familiar with standard deep learning frameworks.
• Familiar with standard machine learning techniques and applications in non - vision domain.

PERSONAL:

• Green Card Holder
• Creative and critical thinker
• Passion in computer vision, machine learning and artificial intelligence
• Fast learner and strong problem solving ability
• Good communication skills

SKILLS:

C/C++, Perl, MATLAB, Python, experienced in Java

Caffe, Tensorflow

AWS Development and Deployment Experience

OpenCV, PCL, VTK, CGAL, ImageMagick, etc.

Linux shell scripts and development tools

Svn, Git/gerrit, SQLite

WORK EXPERIENCE:

Confidential, Chicago, IL

Lead Engineer

Responsibilities:

• Conduct research, prototype algorithms, implement and deliver production ready software for automatic traffic sign detection/classification from Image sequences and LiDAR point cloud data for both core map and high definition map generation for autonomous driving.
• Main contributor to the deep learning based traffic sign detection algorithm, responsible for detector model training, performance evaluation, etc.
• This detector is used to create map content database on a very large scale (billions of images) for a large number of countries.
• Main contributor to the 3D sign localization algorithm that generates localization objects for Highly Autonomous Driving Map Database.
• Research and development of classic computer vision pipelines for 2D image data for automatic and semi-automatic detection of traffic sign/light.
• Coding in C++, Perl, Python, JavaScript, Java
• Provide technical consulting to the management team and coach team members in the areas of computer vision and machine learning.
• Continuously investigate and improve the performance of the algorithm, perform root cause analysis of software and system defects.

Confidential, Natick, MA

Sr. Software Engineer

Responsibilities:

• Conduct research to develop new algorithms for supporting the IDMax and IDQuick barcode finding and decoding technologies.
• Participated in the development of the Humingbird, the first barcode product in the world at the time using the VSoC (Vision System on Chip) technology.
• Designed and evaluated the novel 1D/2D barcode finding algorithms specifically for VSoC.
• Improve the performance of the existing 1D/2D barcode finding and decoding algorithms for the Dataman product.

Confidential

Research Assistant

Responsibilities:

• Developed a general theory for modeling object appearance in an image, learning methods for image appearance models, and explored its applications in object tracking, face recognition, illumination estimation/relighting, super resolution and video compression.

Confidential, Niskayuna, NY

Summer Internship

Responsibilities:

• Proposed a novel convex score function for local minimum free face alignment.
• We compute analytical properties of the score function to enforce the convexity of the surface.
• Independently finished one project that predicted the probability of getting myocardial infarction based on the genotype data.
• Co-worked on another project predicting the outcome of the breast cancer therapy.

Confidential

Internship

Responsibilities:

• Involved in designing the redeye detector and correction algorithm. Improved the hit rate by 5 percentile while keeping the same false alarm rate.
• Theoretical basis for modeling the Image Appearance of an object: We developed a unified mathematical framework for modeling the image appearance by integrating object motion, shape deformation, surface reflectance properties variation, lighting conditions and camera parameters. We also showed how to learn theoretically valid image appearance models by combining the analytical and statistical methods.
• Efficient tracking under varying illumination: Object tracking under large illumination variation continues to be one of the difficult problems. We proposed an efficient reliable and novel 3D inverse compositional method for object tracking under gradual and drastic illumination changes.
• Video-Based Face Recognition: By combining the image appearance theory and the efficient tracking algorithm, we proposed a video-based face recognition algorithm that is able to handle large pose and illumination variations.
• Object-Based Video compression: We compressed the videos of specific objects by building the model of the object, tracking it in the video, and re-synthesizing the video frames.
• 3D Facial Texture Super-Resolution: In super-resolution problem, registration and super-resolution are coupled together, but traditionally they are handled separately. We proposed a closed-loop framework for simultaneous registering and super-resolving 3D facial textures to improve the accuracy.
• Machine Learning in Bioinformatics: Applied a variety of machine learning algorithms and the biostatistics (Chi-2/K-S/T test, Hardy-weinberg equilibrium test) for analyzing the pattern in DNA sequence and predicting the probability of a patient of getting diseases. This work was done as a summer internship in Siemens Corporate Research.