01. About Me
I'm a Full-Stack Engineer based in San Diego who can build scalable software systems on the cloud whether its web apps or micro-service.
I graduated at UC San Diego with a B.S in Computer Science w/ Specialization in Bio-informatics. The degree name is fancy talk for I studied machine learning
and statistics within the biological domain along with software engineering.
I play with a wide variety of technologies around different ends of the tech-stack.
Front-End
-
Javascript -
Typescript -
Angular -
React -
jQuery -
rxjs -
d3.js -
cytoscape.js -
HTML5/CSS3 -
SASS
Back-End
-
Node.js -
Python -
Flask -
Django -
Django Rest Framework -
Celery -
Jupyter -
Conda -
PyTorch
Data Storage
-
Redis -
ElasticSearch -
PostgreSQL -
MySQL -
MongoDB
Dev Ops
-
AWS -
Google Cloud -
Docker -
Kubernetes -
NGINX -
Travis CI -
Github Actions
02. Where I worked
Senior Full-Stack Engineer @ Technical University of Denmark
Dec-2019 to Aug-2020
Prototyped and productionized a web tool for curating visual network graphs of biological & chemical entities along with their interactions/pathways from large public datasets like NCBI, UniProt, & PubMed and vast research literature.
- Utilized vis.js, d3.js, & RxJS to build a HTML canvas drawing tool for constructing interactive network maps of the pathways/relationships between biological/chemical entities.
- Improve REST API performance by shifting compute intensive tasks from web server worker threads to task queue using Celery.
- Refactored Angular codebase from tightly integrated UI elements that were needlessly duplicated many times in the app with inconsistent styling into a re-usable and composable UI components kit built on top of Bootstrap.
#Angular
#D3
#vis.js
#Bootstrap
#Flask
#Celery
#GoogleCloud
#Docker
Full-Stack Engineer @ Re-Hinged
July-2018 to Nov-2019
I led the technical development of web apps that utilized data visualizations to communicate business insights and market intelligence.
- Reduced cloud bill for web apps by 30% through refactoring cloud architecture from monolith to that of micro-services. Used AWS Elastic Container Service to orchestrate containers and auto-scale appropiately on reserved and spot instances with discount pricing.
- Built CI/CD pipeline to handle continuous merging and shipping of new app features between multiple developer teams of different time zones while also ensuring few or no bugs escape to production with checks such as unit testing, code linting, and successful compilation.
- Utilized D3.js and Highcharts.js to build data-driven dahboards that are integrated with NLP pipelines to visualize market sentiment reactions around entities such as brands, products, or technologies.
#Angular
#D3
#Highcharts.js
#Django
#ElasticSearch
#AWS
#Docker
Product Engineer @ Guardiome
Apr-2017 to Apr-2018
I pivoted product development of a clinical genome/transcriptome explorer to a collection of omics clinical analysis apps built on top of Omics SDK.
- Collaborated on Omics Software Development Kit's design/implementation which allows developers to build clinical analysis reports around an user's genome and transcriptome.
- Refactored python bioinformatics pipelines into distribute apps developed on top of the Omics SDK.
- Re-designed the UI and impleneted it with Material Design UI components library to allow easier information digestion of genetics insights and analysis.
#React
#Material-UI
#Electron
#Flask
#Jupyter
#Conda
Software Engineer @ KnuEdge
Apr-2016 to Apr-2018
I contributed to the software development of a machine learning web app which allowed end-users to build machine learning algorithms on proprietary data.
- Implemented visualization components such as line charts on top of D3 and web socket for real-time data streaming/visualizing on top of Angular framework.
- Integrated PlumbJS library into a web app to build a drag & drop interface for creating graph networks that represent machine learning algorithms.
- Prototyped REST API service w/ Node.js/Express & Mongoose/MongoDB.
#Angular
#Material-UI
#PlumbJS
#D3.js
#Node.js
#Express
#MongoDB/Mongoose
03. Projects
David Attenborough Text 2 Speech Synthesis
I took Nvidia's implementation of Tacotron2 to train a deep neural network over 20 hours of labeled audio transcript data of David Attenborough to generate speech audio in very high likeness of his voice.
Data
In order to generate his voice model for speech synthesis, I needed a large amount of labled voice data that is clean, audible, and free of foreign noise artifacts. I extracted a lot of quality audio sample of David Attenborough's voice from audiobooks that he narrated.
In order to generate labeled data from the audio samples, I used PyDub in order to detect silence long enough in the audio samples to split over into smaller audio pieces. I then used QuartzNet model from Nvidia's NeMo Conversational AI toolkit to transcribe the audio speech into text.
Training
I trained it on a 4-GPU instance at Lambda GPU Cloud for $1.50 per hour.
The training was done over a week in sessions where after each session I monitor the progress of the training and tune the hyper-parameters to improve the loss curve. Training in total took a total of around 80 hours to get to this level of voice quality.
04. Contact
Get in touch
Currently in the search for software engineering opportunities either in San Diego or remote.
Contact