Dave Stevens Email & Phone Number

Sammamish, Wa, Us

Bellevue, Washington, Us

Cupertino, California, Us

Owned one of the Siri tentpoles, responsible for leveraging Siri technologies to support partner teams and features across the whole of the Apple product line. Projects include, but are not limited to, large scale deployments of services for multiple internal customers using the latest ML based technologies.Designed and developed the infrastructure behind the AI narrated audio books.

Redmond, Washington, Us

Managed a team of 12 developers and ML experts.Ink Recognition Cognitive Service• Developed Microsoft’s first ink recognition cloud service with full support for ink analysis and over 60 languages.• Achieved O365, SOC and FedRamp compliance for the service with particular attention to end-user data security.• Service in production with Powerpoint already onboarded. Converting the service to a full Azure cognitive service (currently in public preview)Improved handwriting recognition• Extended recognizers to cloud service with adjustments for different constraints and opportunities• Significantly improved the development and evaluation of new recognizers.• Language expansion to newer scripts such as Vietnamese and HindiHighly successful collaborations across Microsoft to bring OCR, Math recognition and improved Chinese handwriting recognition to the Windows platform and Cloud service. Worked with PowerPoint and Whiteboard teams to bring ink intelligence to their respective applications. Continued to work with OneNote both with client side recognition and cloud based recognition to support all platforms and devices.Excellent management and interpersonal skills with 100% approval from my team on the most recent Microsoft Poll for “Diverse & Inclusive Work Group Climate”.Drove hiring for the group with many successful internal and external hires from both college and industry.

Redmond, Washington, Us

Windows TabletPC team Senior SDESelected Projects: • For Win7 re-architected the 8 latin script DLLs into a single handwriting recognizer DLL. Expanded on this to allow for support for 22 languages in Win7. Existing architecture with one large DLL per language was unsustainable with the planned expansion of languages. • Moved the recognizers to language packs reducing the on disk size of a default Windows install by up to 4% during Beta. • Custom dictionaries for handwriting recognition on Win7. • Consolidated the language model building into a single framework under the Windows build environment. • Wrote the Input Personalization framework for Vista to collect user data and improve handwriting recognition accuracy. • 30% speed up in the core handwriting recognition engine with 95% fewer memory allocations.Sustained Responsibilities: • Dev representative for the performance triad in UIP. Responsible for performance goal setting, tracking within UIP. • Manage the local builds and SD integrations. Work with our Serbian sub to provide support for their specific build/branch requirements. • Run daily accuracy verification tests on all recognizers (similar to a very large suite of unit tests). Follow up with any unexpected changes in performance, accuracy/speed/size.

Redmond, Washington, Us

Speech Recognition teamSDE II • Rewrote the language modeling adaptation code in the speech recognizer to make it faster and more robust. Worked with MSR to implement several new ideas to improve the performance of the adaptation. Added several enhancements of my own including relevance and aging of the data to the language modeling adaptation algorithms. • Developed low latency speech detection and noise robustness code/algorithms for the Microsoft Speech Server to improve response time. • Responsible for microphone specification recommendations for SR to Longhorn and Pocket PC. • Wrote an Audio Quality evaluation tool to better assess the databases we have. This is now used extensively by both the test team and in our data collection efforts. • Developed the compressed cepstral coding format for multimodal devices based on research from MSR. Created a version for the Pocket PC which uses 10% of the CPU and less than 30kB of memory.Dev Lead position5/2000 – 7/2002 • 18 months in a Development Lead position responsible for developing the new signal processing end of the speech recognizer and integrating it with the speech recognizer and the training tools suite. Created and integrated the initial version of the new front-end to both the engine and training tools one full milestone ahead of schedule.

Noise compensation algorithms to help improve speech recognition accuracy under noisy conditions.

Us

Selected projects and responsibilities included: • Jun 98 – Mar 99: Technical manager working with MoD contractor responsible for providing speech recognizer for the Eurofighter/Typhoon. Principle researcher for improving the capabilities of the recognizer under extreme flight conditions. • Nov 97 – Aug 98: Technical manager for trials and implementation of speech recognizer in main battle tank. • Aug 96 – Nov 97: Applied speaker recognition techniques to automatic language identification. • Feb 96 – Sept 96: Project Manager and lead developer on “Automatic Speaker Identification and Verification” Project. • Nov 95 – Mar 96: Speech Recognition for Air Traffic Control Systems. • Feb 95 – Nov 95: Research on Speech Recognition in adverse environments.

Ph.D., Non-Linear Prediction For Speech Processing

B.Eng.(Hons), Electronics With Computer Science