Feynman Tsing-Yang Liang, Developer in Cambridge, United Kingdom
Feynman is available for hire
Hire Feynman

Feynman Tsing-Yang Liang

Verified Expert  in Engineering

Machine Learning Developer

Location
Cambridge, United Kingdom
Toptal Member Since
February 19, 2016

Feynman is a product-oriented, full-stack engineer with extensive experience building production web applications and machine learning systems. His disciplined engineering process draws from his extensive industry experience and his leadership in open source keeps him on the cutting edge of modern big data technologies.

Machine LearningScalaReactMeteorFluxReduxSpark SQLApache SparkRDBMSSQLPythonLinuxJavaMongoDBR

Portfolio

Databricks
Python, NumPy, Pandas, Scikit-learn, JavaScript, D3.js, React, Scala, MLlib...
Bridgewater
RStudio Shiny, libsvm, Weka, R, D3.js, Scala
Google
BigTable, NoSQL, Flume, MapReduce, Python, JavaScript, Java

Experience

Machine Learning - 6 yearsPython - 5 yearsJavaScript - 5 yearsScala - 3 yearsD3.js - 2 yearsReact - 2 yearsMeteor - 2 yearsApache Spark - 1 year

Availability

Part-time

Preferred Environment

React, Meteor, Spark, Scala, Vim Text Editor, Tmux, Linux

The most amazing...

...thing I've done is to take 130 massive open online courses (MOOCs).

Work Experience

Machine Learning Engineer

2015 - 2015
Databricks
  • Managed and reviewed open source contributions to Apache Spark MLlib, a distributed machine learning library.
  • Built interactive D3 visualizations for machine learning model visualization.
  • Implemented distributed variants of multiple machine learning algorithms, including: LDA, GMM, and PrefixSpan.
  • Implemented performance testing and QA for the Spark 1.5 release.
  • Contributed low-level optimizations involving off-heap JVM caching to Spark SQL's Project Tungsten.
Technologies: Python, NumPy, Pandas, Scikit-learn, JavaScript, D3.js, React, Scala, MLlib, Spark

Technology Associate

2014 - 2014
Bridgewater
  • Built a D3.js-based decision tree training, analysis, and inspection web app for use by financial analysts and managers.
  • Built a predictive model of employee performance from various data unstructured (e.g. emails, performance reviews) and structured (e.g. telemetry, trading data).
  • Authored a technical report about model uncertainty and confidence which proposed a confidence-weighted averaging scheme now used in production deployments.
  • Performed network analysis of internal company communications (emails and instant messages), identified cliques within company and isolated teams.
  • Proposed and implemented an algorithm for determining next best splits in a decision tree as well as a method for evaluating the value of additional information.
Technologies: RStudio Shiny, libsvm, Weka, R, D3.js, Scala

Software Engineer

2013 - 2013
Google
  • Built a change history data workflow which performs aggregation on more than 100 TB of daily change data.
  • Introduced advanced analytical tools for online advertising campaign management, including segmentation and anomaly detection.
  • Improved the test coverage of Google Adwords front-end by writing unit (JUnit, Mocha) and integration (WebDriver) tests.
Technologies: BigTable, NoSQL, Flume, MapReduce, Python, JavaScript, Java

Software Development Engineer

2012 - 2012
Microsoft
  • Automated telemetry and reporting for Microsoft Office's internal build and test tools.
  • Deployed SCOM/SCCM for managing configuration of entire Office Shared Services multi-device build systems.
  • Built web application integrated with SCOM to display health of build and test services and notifications of outages for internal developer stakeholders.
Technologies: System Center Configuration Manager (SCCM), SCOM, ASP.NET, C#

Large Scale Topic Modeling: Improvements to LDA on Spark

Implemented online variational inference for LDA on Apache Spark, resulting in 20% improvement over original EM implementation.

Improved Frequent Pattern Mining in Spark 1.5: Association Rules and Sequential Patterns

Collaborated with researchers from Huawei to implement a novel distributed variant of a frequent sequential pattern mining algorithm (PrefixSpan).

Visualizing Machine Learning Models

https://databricks.com/blog/2015/10/27/visualizing-machine-learning-models.html
D3.js and React visualizations of various machine learning models, including: linear regression, logistic regression ROC, and k-means.

Shopping with Friends

Real-time (WebSockets) shopping list application which supports notifying/adding items via SMS (Twilio) and payments (Venmo). Second prize at HackDartmouth.

Docuflow

Uses EEG measurements (Muse headband) and pupil-tracking to determine which parts of documents users pay attention to and which parts did not receive much attention. Won the People's Choice award at HackCambridge.

Languages

JavaScript, Scala, SQL, Python, Java, R, C#, Ruby, C++, C, Haskell

Frameworks

Redux, Flux, Apache Spark, Ruby on Rails (RoR), Spark, RStudio Shiny, ASP.NET, Akka

Libraries/APIs

React, D3.js, Node.js, MLlib, Scikit-learn, Pandas, NumPy, libsvm, RxJS, Twilio API, PayPal API, Instagram API, Twitter API, Facebook API

Tools

Spark SQL, Tmux, Vim Text Editor, Weka, System Center Configuration Manager (SCCM), Flume

Paradigms

Functional Programming, MapReduce

Platforms

Meteor, Linux, Apache Kafka, Amazon Web Services (AWS)

Other

Machine Learning, SCOM, Bitcoin

Storage

RDBMS, MongoDB, NoSQL, BigTable, Cassandra, Memcached

2015 - 2016

Master of Philosophy in Machine Learning

University of Cambridge - Cambridge, UK

2012 - 2015

Bachelor of Engineering in Electrical Engineering

Dartmouth College - New Hampshire

2010 - 2014

Bachelor of Arts in Math

Amherst College - Massachusetts, USA

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring