Hands-on:
- C
- Data structure & Algorithm (machine vision)
- Linux & Android (HAL, binder, native library, driver)
- Arm & 8051 base embedded system
Neural network:
- Python, TensorFlow, Numpy, Pandas
- Android Things (NNAPI runtime, driver)
Also :
- C++, java, script(shell, makefile, Android.mk)
- Linux & Android (App, SELinux, trustzone)
- ARM & 8051 SoC
- Soldering
- OrCAD
- PADS
- Circuit concept
- Communication protocol
2021/01/04 ~ 2021/06/30
Description:
To exhibit potential customers their website could be attacked, we build up different login-automation method.
Include Appium, web request, Selenium.
Result:
Successfully dodge defence to auto-login.
2019/06/24 ~ 2020/05/16
Description:
Establish formal process to evaluate each sensor & determine third party.
Break down to each algorithm of "raw image", "image processing algorithm", "enroll algorithm", "verify algorithm", "file compress algorithm".
Result:
The result trend is similar to Goodix which is the largest market share fingerprint sensor manufacture.
2017/07/20 ~ 2018/03/31
Member:
9 members team (COO, a consultant, a manager, me & 5 members)
Description:
Currently, fingerprint sensors are working under glass, touch, and panel with different refractive-indexes and will receive noise inevitably. Therefore, we were trying to understand the real images behind noise with neural network.
Referring to GoogleNet & ResNet, analyzing Tensorflow framwork, interpreter & protobuf, Android NNAPI, we were trying to design a CNN and break it down to the cooperating hardware team.
Result:
It is still on going.
2017/07/01 ~ 2018/03/31
Member:
7 members team (CTO, a manager, me as project leader, 3 members, and a QA).
Description:
Replacing "minutiae" with "feature-matching", CTO recruit us for a new generation algorithm. Deal with images (denoise, normalize, blur). Locate features with Sobel and match.
And a patent to swipe enroll. A patent to avoid fake-fingerprint-invading reported by news.
Result:
Still keep optimizing on speed, space efficiency and conquering special case (humid or dry finger).
This algorithm is now running on Samsung Galaxy S8, S9, and many other cell phones like Sony.. etc.
2015/10/18 ~ 2015/10/29
Member:
3 partner team (with two friends in "Linux & Android" class in cram school).
Description:
It was graduation topic built on embedded-Linux on Arm-Cortex-A9. An USB-Webcam video input. Recognize human face with OpenCV, calculate the coordinates and camera angle to shift in Linux application. Then, control I2C-stepper-motor to shift the camera with Linux driver.
Result:
The result is now demonstrated in the cram school, and I got another course for return.
Temp test engineer for Google mobile phone security test assigned by IMC
Duties:
- Pixel security chip test
- Test automation scripts (mostly Python, Shell script)
Main skills:
- Android phone
- Python, shell scripts
- Pixel security chip, TA
Dynasafe is providing network secure system.
Duties:
- Android framework SDK
- Website attacking methods survey
- Automatic login
- Defend core
Main skills:
- Android app
- Java, Python, C, C++
- Chromenium, Sikuli, Request for web
VIVO is building Android smart phone.
Duties:
- Biometrics sensor tuning.
- Algorithm for fingerprint sensor image value.
- Android system. HAL, HIDL
- Trustzone
- Linux driver
Main skills:
- Algorithm
- C, C++, java, script(shell, makefile, Android.mk)
- Linux & Android (App, HAL, binder, native library, SELinux, trustzone, driver)
- Arm base embedded system
HTC is building Android smart phone, virtual reality device.
Duties:
- User behavior algorithm depends on sensors
- Android system
- Trustzone
- Linux driver
Main skills:
- Algorithm
- C, C++, java, script(shell, makefile, Android.mk)
- Linux & Android (App, HAL, binder, native library, SELinux, trustzone, driver)
- Arm base embedded system
Egistec is a emerging IPO company with about 200 employees for now. Products are mainly fingerprint sensors and algorithm. I was recruited as Android system engineer at first and join algorithm team later on July, 2017.
Duties:
- Speed up & optimize present algorithm
- Realize algorithm from public research
- Android phone project leader of Hisen China
- Design & maintain Software Development Kit for Android phone producer
- Import new Android version
- Import development environment and trustzone from each MCU provider
- Linux driver
- New project of Android-phone combining embedded system
Main skills:
- Algorithm of machine vision
- C, C++, java, script(shell, makefile, Android.mk)
- Linux & Android (App, HAL, binder, native library, SELinux, trustzone, driver)
- Arm base embedded system
ICT is the largest currency-verifying company in Taiwan with about 250 employees. Products are mostly the money-inserted part on vending machine and gambling machine.
Duties:
- Control 8051-base MCU
- Control motors, light sensors and software flow
- Design their own communication protocol base on UART, I2C and SPI
Main skills:
C, 8051 & ARM base MCU, Communication protocol
Syntek is public company with about 110 employees. It's a MCU design house of mostly automotive electronics. E.g. navitgation, recorder.
Duties:
- Design circuit and PCB
- Correct and maintain demo-board
- Design and build up customized hardware solutions
- Realize USB , HDMI, SDcard, display connector or reduce noise....etc.
Main skills:
Soldering, Circuit concept, OrCAD, PADS
-「海洋大學創業競賽」(NTOU new business competition) 2nd place as team leader.
-「NEW BIZ全國大專盃創業競賽」(NEW BIZ competition) 4th place as team leader.
- Designed ICT's own communication protocol.
- Face-tracing cam (base on Linux, OpenCV) within 3-members-team.
- Project leader of fingerprint sensor build-in to Hisense.
- Project of SELinux in Egis.
- Project of Linux driver in Egis.
- Project leader of fingerprint sensor bridge base on ARM m3.
- Chef of welfare-committee in Egistec.
- Chef of reading group in Egistec.
- Principal engineer of fingerprint algorithm on embedded system.
- Patent of Fixed-Pattern-Attack (Project to prevent fake finger).
- Project leader of Android system of NPU.