Chanmann Lim

CHANMANN LIM 7984 21e Av Montreal QC H1Z 3S8 +1 - github://lchanmann twitter://lchanmann linkedin://chanmannlim medium://lchanmann - SUMMARY Winner of the prestigious Fulbright scholarship in 2014 leading to a 2-year master’s degree specializing in Machine Learning and Speech Recognition at the University of Missouri-Columbia, U.S. Through his masters thesis, Chanmann has contributed to the development of speech and language technology for his native language as well as advancing the field of low-resourced speech recognition. The preliminary result of Deep Neural Network based acoustic model trained for Khmer speech recognition in his work performed at 93.13% word accuracy. He has also 8 years of experience designing and implementing start-of-the-art software solutions across multiple platforms, languages and environments. Passionate code craftsman with solid knowledge and expertise in test-driven development, design patterns and Agile software development, he demonstrates superior coding skill by writing highly-testable code (up to 98% test coverage). Challenge-driven geek capable of rapidly adapting to new technical environments; able to work proficiently in both independent and collaborative team environments. EDUCATION MS. Computer Science University of Missouri-Columbia, Columbia, MO GPA 3.875 AWARD/HORNOR – Fulbright scholarship THESIS – Towards Deep Learning on Speech Recognition for Khmer Language BSBA. Management of Information System National University of Management, Phnom Penh, Cambodia May 2016 June 2006 PROFESSIONAL EXPERIENCE InSTEDD Innovation Lab Southeast Asia Software Engineer September 2008 - June 2014 Phnom Penh, Cambodia Achievements: • Played as scrum master to oversee and boost team commitment. • Performed test-driven development, pair programming and maintained high test-coverage. • Built and customized platform products according to unique demands for local and regional clients. • Automated application deployment, data migration and backup to streamline service delivery. • Monitored server resource consumption and alert to detect anomalies. • Assisted recruitment and job training for junior developers. • Increased visibility of the organization and its products by presenting in tech events. • Contributed to open-source (https://rubygems.org/profiles/mann). Projects: • Resource Map (http://resourcemap.instedd.org): allowed agency users in the field to geographically map resources and have access to real time information via SMS regarding resource availability and needs. L. Chanmann p. 2 Technologies: Rails (v3.2.17), MySQL, Apache, Git, KnockoutJS, ElasticSearch, Redis and Nuntium - SMS platform (http://nuntium.instedd.org). • TB-Lab (https://bitbucket.org/ilab/tb-lab): enabled health centers to send information about suspected Tuberculosis (TB) patients and get real-time SMS updates on lab test results. Technologies: Rails (v3.1.10), MySQL, Apache, Mercurial(hg) and Nuntium. • SMS Appointment Reminder: enabled home-based care workers to track and remind HIV/AIDS patients to attend appointments, reducing the number who drop out of treatment. Technologies: C#, Winforms, Windows Service application, SQL Server Express and Java2 Micro Edition (J2ME). ButterfulyMx April 2013 - February 2014 Ruby Developer Upwork (previously known as oDesk.com) • Implemented RESTful API for revolutionary Intercom system. • Set up, deployed, maintained and monitored production server to ensure smooth business operation. Technologies: Rails, MariaDB, Nginx and Git. Indochina Research January 2007 - August 2008 Programmer Phnom Penh, Cambodia • Developed web-based file sharing system to handle the delivery of over-sized client reports. • Developed internal web-based staff and contact database for corporate contact lookup and management. Technologies: C#, ASP.NET and Microsoft Access SilkRoad Cambodia Programmer • Developed internal call center database for call analysis and reporting. • Developed database-driven dynamic website for clients. June 2006 - December 2006 Phnom Penh, Cambodia Technologies: Classic ASP and Microsoft Access. TECHNICAL SKILLS Languages: Ruby, JavaScript, Java, Matlab, Shell. Web: Rails, RESTful, JSON, HTML, CSS, Haml, Yaml, jQuery, KnockoutJS, GoogleMap API. Tools: Vim, git, Mercurial (hg), RSpec, Jasmine, Unicorn, Capistrano, Heroku, Amazon EC2. Servers: Apache, Nginx, MySQL. Platforms: Ubuntu, CentOS. ACADEMIC PROJECTS Towards Deep Learning on Speech Recognition for Khmer Language • • • • May 2016 Master’s Thesis Research Implemented Voice Activity Detection algorithm for data pre-processing. Applied Decision Tree algorithm for triphone state clustering. Trained acoustic models using Gaussian Mixture Model and hidden Markov Model (GMM-HMM). Trained Deep Neural Network (DNN-HMM) hybrid models with Dropout and transfer learning. L. Chanmann p. 3 Technologies: HTK 3.5 beta, Shell, Matlab, Ruby and C. Unsupervised Approaches for Khmer Word Acoustic and Speaker Discrimination • • • • December 2015 CS 8735: Unsupervised Learning Extracted MFCC feature from raw audio files. Extended Dynamic Time Warping to account for warping path normalization. Implemented clustering algorithms such as K-medoids, Fuzzy C-medoids and Spectral Clustering. Implemented Normalized Mutual Information metric for clusters evaluation. Technologies: Matlab and C. Classifiers Comparison for Gender Identification from Facial Images December 2015 CS 8725: Supervised Learning • Extracted and reduced feature dimension using Principal Component Analysis. • Applied Naive Bayes, K-Nearest Neighbors and Support Vector Machines classifiers and k-fold cross validation for image classification. • Implemented AdaBoost algorithm to improve classifiers performance. Technologies: Matlab. An STE-based Methodology for Constructing Bayesian Belief Networks as Applied to the Prediction of Economic Indicators May 2015 CS 8750: Artificial Intelligence 2 • Constructed Bayesian networks according to STE scores and domain knowledge. • Implemented Maximum likelihood parameter learning with Laplacian smoothing. • Implemented Bayesian network inference algorithm using variable elimination method. Technologies: Matlab. Path Finding Problem December 2014 CS 7750: Artificial Intelligence 1 • Built graphical game map with grid structure. • Implemented 6 search algorithms to find path between 2 points on the game map. • Algorithms: Breadth-First Search, Depth-First Search, Greedy Best-First Search, A* Search, HillClimbing, and Simulated Annealing. Technologies: Java and Swing. CONFERENCES & EVENTS • Epihack, Chiang Mai, Thailand (March 2014) Participant – Designed VoIP-based workflow prototype for animal health reporting system. https://www.youtube.com/watch?v=D97aVtLR0rE • TechCamp Southeast Asia, Phnom Penh, Cambodia (September 2013) Speaker – Trained and facilitated brainstorming on how to use open-source technologies to amplify social works. https://techcampphnompenh.wordpress.com/trainers L. Chanmann p. 4 • Open Cambodia, Phnom Penh, Cambodia (September 2011) Participant – Learned and facilitated group discussion. http://www.opencambodia.net • Cambodia first Startup Weekend, Phnom Penh, Cambodia (September 2011) Winner team – Developed SMS-dating application prototype. http://www.up.co/communities/cambodia/phnom-penh/startup-weekend/214 TALKS Building command-line app with Thor. Ruby Meetup, Phnom Penh, Cambodia (2012) Building scalable SMS-based app with Nuntium. Barcamp Phnom Penh 4 (2011) CoffeeScript: JavaScript in a simple way. Open TechTalk, Phnom Penh, Cambodia (2011) Test-driven development and J2ME. DevCamp, Phnom Penh, Cambodia (2010) REFERENCES Available upon request.