williamchan@cmu.edu

William Chan

Career Objective: Machine Learning Specialist with applications in Ads, Social Networking, Finance, Computational Biology, and/or Hardware Design
Career Achievements: 6 Internships: Google, Intel, NVIDIA, AMD, TD Securities, Amazon
Bachelor's:
Sept 2006 - April 2011
University of Waterloo
BASc Computer Engineering and Management Sciences
Jan 2010 - April 2010
National University of Singapore
Computer Engineering Exchange Student
Master's/PhD:
Sept 2011 - Current
Carnegie Mellon University
MS/PhD Computer Engineering (Machine Learning)
Intel - Platform Architecture Engineer (Santa Clara)
May 2008 - Aug 2008
  • Adaptive Power Algorithms Research
    • designed and prototyped several dynamic frequency algorithms for CPU/GPU to reduce dynamic power
    • +55% in CPU power reduction with no performance penalty in certain apps/games
  • Multi-Chip Package Power Controller Integral Algorithm
    • New methodology to increase CPU and GPU performance while maintaining same thermal design power envelope (patent pending)
  • Statistical Data Analysis on Power Traces - analyze and isolate power events for simulation and forecasting next generation mobile platform architectures
  • Added Business Value: additional product value with zero capital and marginal unit cost
Google - Software Engineer (Mountain View)
Jan 2009 - Apr 2009
  • Research and analysis on disk latency and bandwidth
    • Data analysis on disk performance across entire Google's server fleet
    • Designed and implemented new disk benchmarks
  • Added a new Linux Kernel Power Capping Module to output detailed CPU usage to assist in power management across Google's server fleet
  • Added Business Value: maximize utility, lower capital and operating costs, negotiation leverage with disk vendors
TD Securities - Quantitative Financial Engineer (Toronto)
Sept 2010 - Dec 2010
  • TD Bank's trade floor in the Quantitative Research Group (Global Equity Derivatives)
  • Optimized Monte Carlo Simulation through dividend events consolidation
    • On dividend heavy deals (i.e. basket options on indicies), decreased runtime by over 1700% with constant volatility, over 500% with local volatility surfaces
    • Overall decreased trader's live book calculations by around 10-30%
  • Added a new mixed Dividend Pricing Pricing Model
    • Allows pricing of most exotic derivatives with either Analytic, Monte Carlo or PDEs to use a inter-blended variable mixture of discrete dollar and discrete proportional dividend schedule
  • Architected new 2D Interpolation Infrastructure and added Thin Plate Spline Interpolation for Local Volatility Surfaces
  • Cholesky decomposition optimizations for FX/Equity correlation matrices by bootstrapping similar matrices used in Monte Carlo Simulations
  • Added Business Value: faster execution, more accurate pricing, happier clients, better risk management and hedging
NVIDIA - Software Engineer (Santa Clara)
Sept 2007 - Dec 2007
  • Driver Optimizations
    • Various driver optimizations including in +20% performance on Call of Duty 2 on G80
  • Added Business Value: additional product utility without additional capital or marginal costs, product differentiation against competitor
AMD - Software Engineer (Markham)
Jan 2007 - Apr 2007
  • HD2Player - High Bandwidth Video Playback Research and Development
    • Researched and developed a custom video system designed for 1920x1080 @ 120Hz playback with throughput of over 6Gb/s and designed to scale over 240Hz
  • DVI2LVDS Board
    • Using Verilog, developed a FPGA for a custom proprietary board for converting a DVI/HDMI signals to LVDS for a variety of LCD panels
    • Solution streamlined the process to use various vendor prototype panels
  • Floating-point Matrix Optimizations
    • Optimized existing matrix code with SSE2 and SSE3, in some cases performance increased over 100%
  • Added Business Value: capital cost savings (> 100k CAD) with custom in-house solutions vs commercial purchases
Amazon - Software Engineer (Seattle)
Sept 2009 - Dec 2009
  • Prime AdAttribution
    • Re-architected database schemas (Oracle and MySQL)
    • Modified data mining infrastructure to more efficiently attribute Amazon Prime signups to specific ads for statistical analysis
  • Created an internal GUI tool for non-technical business users to access and utilize large quantities of customer data
  • Added Business Value: faster and more effective analysis of customer behaviour