Cody Wu

FAQ About Cody Wu

• What is Cody Wu's address?

  Cody Wu is located at
  Palo Alto, California, United States

• What is Cody Wu's job title?

  Cody Wu is a
  Dr. Wu is a top-notch expert in the fields of algorithms, LLM learning, error-correcting codes, data compression and deduplication, and CPU optimization. Dr. Wu has published 20+ world-class journal papers and 40+ US patents across a wide range of AI, computational algebra and graph, error-correction coding, data compression and deduplication, VLSI, etc. Dr. Wu invented the rational-curve-fitting algorithm, dubbed the Wu algorithm, reducing a classic list decoding problem from naively O(2n) to O(n6). Apr. 2013, Dr. Wu Co-founded Tidal Systems which was acquired by Micron Technology in Oct. 2015. 2023-24 Adaptive near-deduplication for training data while retaining integrity. Developed the largest high-performance FFT library (in C++) containing ~150 algorithms. Fine-tuned ChatPGT/Llama2 to code various highly optimized FFTs in C++. LLM also inspired me to devise the most efficient FFTs over all radixes. 2022. Developed end-to-end AI tensor transformation in runtime to nearly saturate memory bandwidth. 2021. Contributed to the prevalent open-source libjpeg-turbo (github.com/libjpeg-turbo) to expedite JPEG decompression by 20~40%. Contributed to the prevalent open-source zlib (github.com/madler/zlib) to speed up zlib/gzip decompression by ~40%. 2019. Devised the first-in-class Rate-compatible QC-LDPC codes to suit arbitrary parity length for SSD. Devised a novel decoding algorithm that uses 4-bit LLR to outperform floating-point normalized min-sum decoding. 2018. Devised a novel deduplication algorithm that enables deduplicating imperfect data chunks. Its complexity is comparable to the existing deduplication schemes based on perfect matches. 2017. Devised an advanced RAID scheme using the integrated-interleaved coding scheme. First-time devised the systematic encoder for integrated-interleaved Reed-Solomon and BCH codes. 2012. Overturned the sweeping conjecture in the past 50 years that binary Reed-Solomon codes are asymptotically good. Devised a one-pass Chase (soft) decoding of Reed-Solomon and BCH codes by reducing each trial decoding complexity to O(n) from O(n2). 2008. Devised a rational curve fitting algorithm that dramatically reduces the list decoding complexity of Reed-Solomon, BCH, and Goppa codes, dubbed as the Wu algorithm. 1999. Developed novel associative memories using neural networks which exponentially increase the storage capacity. Sambanova Systems

• What industry does Cody Wu currently work in?

  Cody Wu currently works in the
  Sambanova Systems

• What is Cody Wu's area of expertise?

  Cody Wu specializes in
  Dr. Wu is a top-notch expert in the fields of algorithms, LLM learning, error-correcting codes, data compression and deduplication, and CPU optimization. Dr. Wu has published 20+ world-class journal papers and 40+ US patents across a wide range of AI, computational algebra and graph, error-correction coding, data compression and deduplication, VLSI, etc. Dr. Wu invented the rational-curve-fitting algorithm, dubbed the Wu algorithm, reducing a classic list decoding problem from naively O(2n) to O(n6). Apr. 2013, Dr. Wu Co-founded Tidal Systems which was acquired by Micron Technology in Oct. 2015. 2023-24 Adaptive near-deduplication for training data while retaining integrity. Developed the largest high-performance FFT library (in C++) containing ~150 algorithms. Fine-tuned ChatPGT/Llama2 to code various highly optimized FFTs in C++. LLM also inspired me to devise the most efficient FFTs over all radixes. 2022. Developed end-to-end AI tensor transformation in runtime to nearly saturate memory bandwidth. 2021. Contributed to the prevalent open-source libjpeg-turbo (github.com/libjpeg-turbo) to expedite JPEG decompression by 20~40%. Contributed to the prevalent open-source zlib (github.com/madler/zlib) to speed up zlib/gzip decompression by ~40%. 2019. Devised the first-in-class Rate-compatible QC-LDPC codes to suit arbitrary parity length for SSD. Devised a novel decoding algorithm that uses 4-bit LLR to outperform floating-point normalized min-sum decoding. 2018. Devised a novel deduplication algorithm that enables deduplicating imperfect data chunks. Its complexity is comparable to the existing deduplication schemes based on perfect matches. 2017. Devised an advanced RAID scheme using the integrated-interleaved coding scheme. First-time devised the systematic encoder for integrated-interleaved Reed-Solomon and BCH codes. 2012. Overturned the sweeping conjecture in the past 50 years that binary Reed-Solomon codes are asymptotically good. Devised a one-pass Chase (soft) decoding of Reed-Solomon and BCH codes by reducing each trial decoding complexity to O(n) from O(n2). 2008. Devised a rational curve fitting algorithm that dramatically reduces the list decoding complexity of Reed-Solomon, BCH, and Goppa codes, dubbed as the Wu algorithm. 1999. Developed novel associative memories using neural networks which exponentially increase the storage capacity.

• Which languages does Cody Wu speak? 

  Cody Wu speaks the following languages:
  No languages available.

• Which industries has Cody Wu previously worked in?   

  Cody Wu has prior experience in the following industries:
  Tenafe , Micron Technology , Tidal Systems , Sandforce , Link A Media Devices

• What is Cody Wu’s work experience?  

  Cody Wu has the following professional experience:
  Sambanova Systems , Tenafe , Micron Technology , Tidal Systems , Sandforce , Link A Media Devices .

• How can I find Cody Wu on LinkedIn?  

  You can view Cody Wu’s
  LinkedIn profile at LinkedIn URL.

• What are Cody Wu’s work email addresses?  

  Cody Wu’s work email addresses are
  [email protected] .

• What company does Cody Wu work for?  

  Cody Wu works at
  Sambanova Systems