Principal Engineer (Project Manager / PMO)
Samsung Electro-Mechanics Co. Ltd., SEM-B
Sensor Data Analysis (SDA) & Deep Learning Portal (DLP): SDA is a recurrent neural network based on live sensor data like temperature, pressure, humidity etc. from embedded devices to judge the status of the stacker machine & predict the probable downtime of the machine in near future. DLP is a Java based portal for customized user specific DL model creation based on sensor data uploaded which could be used for SDA.
- Team Size: 8
- Contribution: DLP product design, System integration & demonstration in 6 months timeline.
- Achievement: 1 month advance early warning on stacker machines – KW 18,000,000 envisaged
- Technologies: Deep Learning, HDFS, Spark, Kafka, MQTT, Hbase, Apache Thrift, Java, Rest API, Tomcat 7, MySQL, Python, JSON & XML
Advanced Data Analysis Management (ADAM): Web-based PCB defect classification and monitoring portal with graphical statistics based on different defects. It also includes an MFC based product to track the defects identified in PCB, remove false alarms using CNN based deep learning (supervised learning) and intimate the visual rectification system. The product was designed to fully automate the PCB manufacturing process with lesser defects and centralize the defect analysis to avoid frequent travelling of engineers to plants.
- Team Size: 8
- Contribution: Product design, System integration & deployment (3 plants) in 1 year timeline.
- Achievement: 60% reduction of false alarms, 30% reduction in Inspection time
- Technologies: Deep Learning, MVC C# .Net 4.5, Rest API, IIS 7, MFC, Java, Java-script, SQL, Windows Services, JSON
Unified Communications (VOIP): Voice over IP based communication platform to improve data exchange between operators, plant managers, HQ experts & IT. The solution involves android and windows based UI to make AV calls, group calls, chats; document & AV attachments, and send status / alarms to the concerned.
- Team Size: 12
- Contribution: Product proposal & design, Quality Control (baseline).
- Achievement: Deployed in 6+ plants and saved KW 40,000,000.
Technologies: (Linux, Android & Windows) Asterisk 12, Rest API, JSON, OpenFire, MySQL, Oracle, LinPhone, Jitsi and Mosquitto