Bo-Yuan Yang
Graduated from the Institute of Applied Physics, National Taiwan University. The field of the master degree's research was experimental high energy physics. I have participated in the Belle project, which is located in Japan. My master thesis research mainly uses machine learning to analyze collision particles decay mode with big data size. Having a strong interest in big data analysis and machine learning. I'm already performed my military service, and work at PEGA corp as the software engineer now.
Email：[email protected]
Phone：0939575005
address：9F, No.22-5, Le Li 3rd Street, Anle District, Keelung city.
Born date ：1991／9／22
Education
- National Taiwan University／Institute of Applied Physics, master degree 2014／09～2017／09
- Advisor: Prof. Pao-Ti Chang
- Research Area: Experimental High Energy Physics
- GPA: 3.98/4.3 (overall)
- Yuan Ze University／Department of Photonics Engineering, bachelor degree 2010／09～2014／06
- Advisor: Prof. Nien-Po Chen
- GPA: 3.16/4 (overall)
- Ranking: 19/52 (overall)
- Taipei Municipal Songshan Senior High School 2007／09～2010／06
Experience
Experience 1，Mar, 2019 -
Pegatron corp. , Software Engineer
- Deep-learning data engineer.
- Platform engineer.
- Kafka: Develop/maintain consumer service (Python) on platform; Tuning consumer performance from 5 msg/s to 80 msg/s; multi-site management of consumer pods on Kubernetes and set-up consumer alert mechanism by prometheus; Develop k8s pods/service/hpa management toolkit to allow developer to modify deployment/service/hpa yaml or update image version more easily on k8s (bash).
- Platform backend develop: Resource/project manager API design/implementation (Python aiohttp).
- Model performance tracking dashboard backend API: Classification confusion matrix, softmax bar chart; landmark box plot; segmentation AP chart and FA table; edge rejudge pass/fail confusion matrix/summary table.
- Elastic search/ Kibana operation.
- Platform file browsing system backend develpy / maintain (The backend API server to communicate with mongo DB to get file meta for file listing ) (Python).
- Solution manager develop / maintain, the API server to let model provider to launch/update model on platform and let customer to buy / clone model on platform.
Experience 2，Sep, 2017 - Feb, 2018
National Taiwan University, General Physics Teaching Assistant
- Assist teachers to perform the pre-examination and after-school tutoring.
- Correct the assignment and the midterm, final exam. Use Latex to make the answer sheet.
Experience 3，Sep, 2016 - Oct, 2016
KEK Belle experiment, Japan・researcher
- Assist in the installation of the Belle II CDC detector (Install the cable lines into the readout board) in Tsukuba, Japan.
- Participate in B2GM conference and discuss my thesis project progress with Belle EWP group conveners.
Experience 4，Jul, 2015 - Sep, 2015
Chian Biomedical Tech / Summer Internship
- Using mad-x (developed by CERN, the source code is written in Fortan77) to calculate the relevant physical parameters of the Synchrotron used in proton therapy (slow extraction in specific magnet configurations).
- Participate in discussions and simulations of the magnets configuration in proton therapy Synchrotron.
Experience 5，Sep, 2012 - Jun, 2014
Yuan Ze University, General Physics Experiment Assistant
- Assisting in advise students to accomplish the experiment.
- Design procdure, content and lecture of experiments.
Skills
Tools／Software / Skills
- Moderate C++／C
- Python
- Git
- Latex
- Linux／Mac OS
- Data analysis procedure -- Including multi-variable analysis, modeling(fitting), systematic uncertainty estimation, ...
- bash
- Mongo DB
- Kafka
- Kubernetes
- Docker
- CI/CD flow
- Elasticsearch / Kibana
- Backend develop
Language
Chinese (Native)
English (Upper-intermediate)
Taiwanese (Native)
PEGA Corp. project
Deep-learning platform for serving smart manufacturing.
Kafka conaumer
- Develop consumer as the micro-service on Kubernetes to consume data (device picture) and manufacture logs from factory to server in TPE. The consumer service include several thread to deal with different channel to ensure performance, there are steamer channel which is the Kafka-steaming service to process the log on memory and re-send the message back to Kafka, the file dump channel which is the channel to process/assemble the segmented picture, the log dump channel which consume the manufacture log to TPE for factory machine debug.
- Multi-site management of consumer on Kubernetes and set-up consumer dead , consumer lag alert from Prometheus.
- Turning consumer performance to ensure processing data real-time (from 5msg/s to 80msg/s).
- Use kafka-python module to implement the consumer service and set-up CI/CD flow on gitlab to auto deploy consumer service to multi-site Kubernetes.
- Implement k8s toolkit in bash to easily update consumer deployment / hpa / service yaml for all consumer deployment on k8s by one command.
- Platform resource mgr : Implement the set site API for user to packing/download edge tool and deploy consumer by given site. Maintain the API to get storage/cpu/gpu resource for client project.
- Platform project mgr: Implement several API for other service to communicate with MongoDB by project to get project information.
- Design API spec and implement backend API for model performance tracking to get/calculate value from machine to machine log to front end to show the performance tracking plot.
- Maintain Elasticsearch/Kibana for internal RD to get client machine to machine log.
- File list : Maintain backend API to allow platform file browser front-end to get file listing meta from mongo, let label platform front-end to get file meta to packing dataset, API service for consumer to update file meta to mongo DB, ...
- Solution mgr : Maintain backend API for model provider to launch / update model on platform.
Skills：Python coding, Kubernetes, Docker, Mechine learning platform architecture, MongoDB, Kafka, Elasticsearch, backend develop, bash
Abstract of M.S. thesis
Title：Search for B → K(∗)νν ̄ with Hadronic Tag at Belle Experiment
This thesis reports the study of the B decay channels B → K(∗)νν ̄ with hadronic tag, the K(∗) refer to K+, K∗+, K∗0 and Ks. This study optimized the signal efficiency comparing with the previous study “Search for B → h(∗)νν ̄” [1], performed the bin-by-bin study on full Sb spectrum, where SB is defined as q2/m2B, q2 is missing mass square and mB is the mass of the B meson. The analysis is studied using the complete data of BB ̄ pairs collected at the Υ(4S) resonance with the Belle detector at the KEKB asymmetric energy electron-positron collider. As there is no significant signal yield, this thesis obtains the center value partial branching ratio for each bin and also set the 90 % C.L. upper limit. For SB < 0.3 region of B(B+ → K+νν ̄) < 1.09 × 10−5, B(B+ → K∗+νν ̄) < 3.16 × 10−5, B(B → K∗0νν ̄) < 1.80 × 10−5 and B(B0 → Ksνν ̄) < 0.81 × 10−5. This thesis set the world best upper limit for B → K(∗)νν ̄.
Skills：C++ coding, Data analysis skills, Root, Machine learning, Histogram modeling.
B.S. Project
Internship in Chian Biotech
Chian Biomedical Tech INC is a medical device solutions provider that design/develop/manufacture the proton-targeted cancer therapeutic machines. Proton therapy refers to use a particle accelerator to target a tumor with a beam of protons. These charged particles damage the DNA of cells, ultimately killing a tumor by stopping their reproduction. The motivation of my project was to simulate the slow extraction efficiency under different lattice configuration. We performed the simulation under MAD-X package, which was developed by CERN. Sextuple in the lattice will occur third-order resonance, with the electrostatic septum, we can slow extract the particle that is accelerated to a specified energy. The final simulation result was consistent with our hypothesis, verifying that the lattice parameter I set was a feasible solution.
Skills：Fortan coding, lattice design, parameters optimization.
The left figure is the magnet configuration of Synchrotron.
The right figure is the phase diagram of Third-resonance.
Opencv project
Autobiography
My name is Bo-Yuan Yang. I majored in Photonic Engineering at the university. Besides learning the basic subjects (electronic circuit, electromagnetics etc... ), I also took courses about semiconductor physics and semiconductor Process. I participated in the project about electromagnetic models simulation using Matlab with the FDTD method. In addition, Under the guidance of my supervisor, I used the embedded system Arduino to design a pet automatic feeding machine and took the second price in the competition which was held by the college.
I worked for the physics master after B.S. I participated in the Belle experiment, an international collaboration of more than 17 countries, 400 physicists and engineers, at the High Energy Accelerator Research Organisation (KEK) in Tsukuba, Ibaraki Prefecture, Japan. Belle experiment is to research the B physics, the B meson pairs are produced by an electron and positron collision. My research is focusing on big data analysis of the B meson decay channels. The data sample is collected by Belle detector, the total number of BB bar pair are up to 771 million. In order to eliminate experimenter’s bias, we perform the blind analysis. We study the signal and background with Monte Carlo (MC) simulation. Unless the analysis is completed, the real data will keep blinding. In my thesis, I used the neuro-network algorithm to process the big data of B meson decay final states to knock out the irrelevant data. After the neuro-network selection, we can get the clean relevant data. With that data, we can use the statistical method to calculate the branching fraction. For the period of time I work in Belle collaboration, I learned not only the knowledge of physics and analysis but how to work and communicate with those people from different country and culture in the group.
In addition, I had the internship during the master's degree. The project of the intern job was about to calculate the physics parameter of the proton therapy Synchrotron, and also participate in the discussion of the configuration of Synchrotron's magnet and run the simulation to verify the lattice design.
I'm now working at PEGA corp. as the platform / data engineer RD for developing and maintaining the deep learning platform.
I'm interested at big data field and semiconductor industry. Particularly, If you feel that I am suited for the job that you have in mind, please inform me of the time convenient for an interview. I hope to hear from you in the near future.
