I worked on developing the embedded software for Truly Wireless Stereo (TWS) Bluetooth earbud based on Qualcomm's QCC512X SoC. I ported many features from existing CSR867X sink application and also wrote a significant amount of audio processing code for QCC512X. Notable jobs:

• Wrote a LPC-based speech bandwidth expansion capability in Kalimba DSP asm with necessary controller logic in C.

• Created reusable & tunable mini-batch K-means clustering pipeline to produce vector quantization codebook used in the above bandwidth expansion using Python ML stack (Jupyter, Numpy, Scipy, Scikit-Learn, HDF5)

• Did a lot of research to find and devise the speech bandwidth expansion algorithm with feasible runtime requirements, MIPS and memory-wise.

• Helped my team implement man-machine interface requirements in C and using Sink Application Configuration Tool.

• Worked on integrating touch and proximity sensor ICs via PIO configurations and I2C controlling logic.

• Ported a noise detection-based microphone switching algorithm written in CSR867X into a new Kymera architecture, writing capabilities (unit of audio process) and controlling logic.

• Developed an iOS library to support communication with the earbud via Bluetooth Low Energy, along with a sample application and documentation.

• I also wrote a python-based GUI tool to speed up the flashing/debugging of devices.

I co-founded a data science startup with a few of my UC Berkeley peers. Here's what I did:

• I lead the development and launched the closed beta of a data-driven fashion commerce web service based on React.js/Redux, Express.js and PostgreSQL. It was deployed on Elastic Beanstalk and we used AWS S3 for assets. We integrated with the Korean PGs to handle payments, and we also developed an CRM web app to track/handle orders and deliveries.

• I lead the development of the alpha version of a near-realtime e-commerce recommendation service based on ELK stack (Elasticsearch, Logstash, Kibana), PySpark, AWS Lambda and AWS Kinesis.

• I lead the development of the alpha version of DeepStudio, a browser-based data wrangling platform based on React.js/Express.js, Django, PostgreSQL and Python ML stack (Numpy, Scikit-Learn, Pandas, etc...)

Experience with lecturing to medium-sized (20~30) group of students, teaching Java and concepts relevant to data structures and algorithms.

Android/iOS, JavaScript library, Node.js/Express.js server development. Developed Android application with Bluetooth Low Energy connection with embedded device and data synchronization between remote server and local SQLite3 database (DAO).

Experience with writing webview-based hybrid application SDK bridging Javascript with iOS/Android native features (Bluetooth, NFC, Camera, and etc). Designed the API conforming to W3C Recommendation/Working Group API's. Managed JavaScript build and dependencies with Grunt.js, Bower.js, and npm. Managed company server's docker scripts/images/containers.

Performed multivariate polynomial regression analysis for estimating body temperature from raw sensor values of a thermometer device.

Experience with writing embedded software for TI MSP430 and CC2541. Wrote interrupt handler for MSP430 to control GPS/GPRS module via 2-wire UART and handle application logic. Implemented Firmware over the Air (FOTA) for MSP430 and CC2541.