Jan 2017 - Present
1.Survey/Design architecture of MLOps and BI for new ML project: energy consumption prediction(2021)
2.Deploy/monitor services( MES, ERP adapter, Camera/speaker anomaly detection ) by writing helm chart, dockerfile and yaml to AWS, Azure and Kubernetes.(2020-2021)
3.Completed POC of smart building management cloud platform by python, ifcOpenShell and Kubernetes. Data format: BIM, cobie and Omni Class and TAICS (2020~2021)
4.Participate in data Standard set-up of TAICS smart building (2020~2021)
5.Participate in Intelligent building Evaluation Handbook of Taiwan Intelligent building Association. (2020~2021)
6.Completed implementation of ERP adapter for MES based on ISA-95 by python, MQTT, kubertnetes and MSSQL(2020)
7.Completed implementation data pipeline for speaker anomaly detection(detection accuracy 99%) by python, kubernetes(2020)
8.Assist III’s analyst to deploy airplane tracking solution to cloud platform for airport in Taiwan(2019~2020)
9.Project Manager of Cloud Platform of intelligent transportation system.(2019)
■ Completed anomaly detection for surveillance based on image processing by python and openCV. Representation by Grafana and influxdb. (2019).
10.Proposal intelligent transportation system and cloud and edge computing of TDP(Technology Development Program) of Ministry of Economic Affairs in Taiwan. (2018).
11.National IoT Platform / WISE-PaaS (2017~2018 cooperate with advantech co. ltd)
■ III side Project Manager of AFS(analytics framework services) which help developer implement their solution on cloud quickly by drag and draw style, running as serverless computing based on cloud foundry (2017~2018)
■ Completed architecture design and implementation of AFS by Python, node-red, cloud foundry, and Microsoft azure. CI/CD by Jenkins and GitLab (2017~2018)
■ Completed edge device system integration with cloud and ML model deployment as MLOps by edgeX, docker, consul and MQTT(2018)
12.Assist Taiwan’s steel company running their motor anomaly detection solution on AFS(2018~2019) P1:30 sensors P2:300 Sensors, P3:3000 Sensors
■ there are 30 sensors on one motor, every sensor get 8192 data/sec
■ By using AFS MLOps to increase model accuracy from round1 (82%) to round3 (97%)