Ramin Rezaiifar

**** ******** ***** ** *****, Del Mar, CA 92014

adcs7c@r.postjobfree.com 858-***-****

Summary

I am a computational Imaging and AI engineer with focus on practical algorithm design for embedded platforms with stringent power and latency requirements. I created the Computational Photography team in Qualcomm in 2011 (prior to that I designed wireless MAC/security protocols). We designed, implemented and commercialized several unique algorithms. I am a hands-on leader and have directly contributed to design and implementation of almost all of the algorithms listed below. After Qualcomm, I started designing an AI-based algorithm for computational tomography and underground salt discovery (more details below).

What I Can Do

I seek problems that can benefit from computer vision and deep learning. I enjoy using state-of-the-art techniques to create solutions to real problems and productizing them.

Experience and Examples of My Work

Aug 2019 -- Present

Overview.ai: AI Designer:

Designed and implemented an AI model to robustly detect rare anomalies in the manufacturing process. Attributes of the design:

- The AI model is able to achieve high degree of precision and recall scores

- The model needs to be trained with only very few examples of anomaly

- Computing the lay-length of wires using computer vision techniques

June 2018 – Dec. 2018

Deep Learning Algorithms:

- Computational Tomography: Developed a system based on deep learning to label MRI image for several organs and create a 3D model of tumors (see example video below)

- The system does not require any human interaction and automatically adapts to different organs and produces state-of-the-art scores (DICE, precision, recall).

- Uses computer vision techniques to refine the results and increase precision/recall (in addition to neural networks)

- Examples:

Link to video: https://youtu.be/bynq2hIxV_0

- Detecting salt deposits below the surface of the earth: I developed an algorithm based on deep learning and CRF and ensemble learning to detect salt underground given seismic imaging. It was necessary to combine several machine learning technique and computer vision techniques to increase the DICE score for segmentation.

- Examples:

Left to right: Seismic image (input), Algorithm prediction, Ground truth

2009 – Present: MaaDoTaa.com (Founder):

- Created the iOS development company and published several apps (~25 apps). Most recent app is for Tesla owners and it has quickly become the favorite third-party Tesla app with ~19000 active paid users.

- I have developed the apps and wrote code for the back-end server, cloud functions, and the front-end iOS/watchOS app.

- “Wall color AI”, AI Selfie”, “Profile Image” and “Pet Selfie” use AI algorithms which run directly on the iPhone/iPad.

[2011-2018] Qualcomm, Sr. Director of Technology Deep Learning Algorithms:

- Portrait segmentation used for single-camera/single-frame bokeh effect (real-time and still image on mobile devices (https://goo.gl/hj1xWW)

- Extremely low-light imaging: Restores color and surpasses noise using AI and fusion techniques:

Left: original image, Right: Output of the algorithm

- Super-resolution: Enhances resolution of low-res images ( click on this link to see the higher resolution image: http://bit.ly/2vCAlq8)

- Precise Pedestrian segmentation (https://bit.ly/2vHjKRY)

- Auto White-balance (AWB) using deep learning

Conventional CV algorithms:

- Fusing mono and Bayer images to reduce noise and increase resolution

(http://bit.ly/2vyU9e1)

- Low light video and still image enhancement and noise reduction (http://bit.ly/2vyU9e1 and https://goo.gl/n1Mot4)

- Fusing flash and no-flash images(https://goo.gl/nvwPhl)

- Tap to change plane of focus after taking the shot (uses ordinary camera with no extra hardware) (https://goo.gl/n1Mot4)

- Ghost-free High Dynamic Range imaging

[1996-2011] Qualcomm, Sr. Director of Technology

- Designed and patented a novel protocol for speeding up http download on wireless networks (fast networks with relatively long round-trip-time).

- Designed and patented a novel method for reducing bandwidth on the reverse link of wireless networks for http.

- Designed and standardized security layer and MAC layer and above (mobility management, data services, authentication, encryption, etc.) for several wireless cellular systems (IS-95-B, cdma2000, EvDO)

Education

University of Maryland

Electrical Engineering: MSc [1993], PhD [1996]

GPA 4.0

Sharif University of Technology (#1 technical university in Iran) Electrical Engineering: BSc. 1989 [19 out of 20; Highest GPA in class] Misc.

“Orthogonal Matching Pursuit: Recursive Function Approximation with Applications to Wavelet Decomposition”, B. Pati, R. Rezaiifar, P.S. Krishnaprasad. ~2700 citations in past year, (https://goo.gl/fqvBaq)

More than 150 issued patents (http://goo.gl/geuTWW)

Hobbies:

o Robotics

o Home automation using Insteon, Homekit, iBeacons o Writing apps (several apps in the App Store related to Tesla, home automation, safe driving)

o Rock climbing, Rowing

Contact this candidate