Stories & Testimonials
A selection of success stories from projects managed by Vypian staff.
Providing a glimpse of how Vypian can help your hi-tech business grow.
I’ve known and worked with Dr. Gary Jamieson for over 20 years, first as one of my software leaders and later, as an executive at my largest customer. I have seen him grow and evolve over that period into a highly effective leader and technologist. Gary is above all a pragmatic and driven problem solver. He can ascertain root issues, determine the right technical solutions and drive execution on complex systems, large or small. He is a skilled leader with the ability to assimilate and deploy technology rapidly.
Gary has managed the development, testing and deployment of large-scale software systems that encompass embedded software stacks in devices to massive cloud computing platforms. He is tenacious, detailed oriented, and success driven. Gary is someone that I trust completely and would unhesitatingly use to implement technology.
Joe Chow
Advisor & Investor
President & SVP, Home Networks, CommScope, 2019-2024
Internet & TV devices deployed on a massive scale
Challenge​
To develop, globally deploy, and maintain Internet gateways, video settop boxes and smart TVs for a Fortune 100 service provider business. Requirements included developing a scalable service platform that would not only provide fundamental services such as Internet connectivity and TV channels but could be used to launch additional revenue-generating services including home security, gaming, in-home medical care, and any number of applications. Time to market of these services was critical and high customer satisfaction was paramount.
​
Solution
As Vice President of Product Development, Vypian’s founder Dr. Gary Jamieson directed teams responsible for the development, deployment, and maintenance of the broadband and video devices that formed the cornerstone of a Fortune 100 business. He created highly productive teams and worked effectively with partners and customers to deliver product to market on schedule and within budget. Drawing on his years of experience driving complex engineering development, he directed multiple specialized teams including software integration, device bring-up, cybersecurity, build tools, and machine analytics. Working adeptly across multiple disciplines, cultures and geographies he collaborated with disparate teams including program management, quality assurance, product management, legal, and operations to deploy flagship devices around the globe.
Result​
Global deployment of over 100M devices forming a scalable, revenue generating platform. Allowed the business to generate an annual recurring revenue of >$80B.
I wholeheartedly endorse Gary Jamieson and the remarkable expertise provided by Vypian in the Video and Broadband sector. With an impressive legacy of 24 years in the broadcast device delivery market, Vypian boasts a wealth of experience that is second to none. Gary’s profound knowledge of one of the industry’s premier global middleware platforms, RDK, stems from his decade-long senior leadership tenure at Comcast. His unwavering dedication to the quality delivery of Video and Broadband devices has been instrumental in ensuring that millions of devices reach consumers and businesses globally. Gary’s relentless focus on meticulous procedures, uncompromising quality standards, fostering consensus, and thorough follow-through has cemented his reputation as a highly successful and trusted leader in the industry. It is without reservation that I commend Gary and Vypian for their substantial contributions and outstanding performance.​
RDK Delivery Director
November 2024
Robust cybersecurity across millions of networked devices
Challenge
A population of millions of devices deployed in customer homes and businesses around the globe had to communicate securely with endpoints and services on a network including servers, back-office services, and content sources. Each network device could potentially communicate with several hundred different endpoints as part of its normal operation. Data security, especially customer related data, was of utmost importance. This level of security is provided by using digital certificates - essentially secrets known only to legitimate parties on the network - that are renewed on a regular cadence to increase security (similar to updating your online banking password every few months). At this massive scale of deployment, the handling of these secrets presents a significant technical challenge that needed to be carefully managed to avoid service outages due to expiration or nefarious activities designed to subvert the secrets infrastructure.
​​
Solution
Vypian’s founder directed a cybersecurity team responsible for the security and data integrity of millions of networked devices around the world, including the management and administration of digital certificates. Solutions developed by the team included:
-
Certificate tracking: Developed tools to dynamically monitor certificate usage on client devices, tracking endpoints, certificate owners, and expiration dates, logging these details to a database that triggered certificate update cycles.
-
Certificate lifetime extension: Where possible certificate lifetimes were extended beyond the expected lifetime of the device. With the ability to revoke long-lifespan certificates if compromised, this action essentially reduced the management effort to zero for the certificates in question.
-
Hardware root of trust: Enabled unique certificates per client device ensuring any compromised certificate would only affect a single client and not multiple clients.
-
Robust upgrade process: Process and governance were established to ensure certificates were updated at least one year before their expiration date.
-
“Time machine" testing: As part of the standard quality assurance test cycle, artificially set the system time to 1-2 years into the future to identify and correct issues related to certificate expiration.
Result
Network security was robust and industry standards compliant. Maintenance activity became simple, repeatable and efficient. There were no outages due to certificate expiration or unauthorized activity.
High availability software build system for global development team
Challenge
To support a global team of >1000 software developers operating 24/7, a robust developer environment was required that managed source control and image builds in a continuous integration environment. With a demand for over 1000 official builds and thousands of debug builds each day, the system had to be intuitive, high-availability, and integrate seamlessly with adjacent infrastructure including software test racks, bug tracking databases and field deployment systems.
​
Solution
Dr. Jamieson managed a built tools and developer experience team that was responsible for the development, maintenance, and continuous improvement of a ‘Platform as a Service’ build system. Key components of the solution included:
-
Leveraging Open-Source packages including Jenkins and Gerrit to orchestrate source management, image builds, and code reviews.
-
A developer-accessible web portal to monitor build status and access software images.
-
Comprehensive monitoring for early detection of system outages and performance issues.
-
A layered build architecture that facilitated software object re-use, reduced inter-component dependencies, ensuring only necessary compilations were triggered and build times minimized.
-
Agile DevOps methodologies were used throughout.
Result
A comprehensive build system that supports a large global team of developers. Through monitoring and redundancy the system is available 24/7 and has the flexibility to evolve as business demands change. Significant advances to date include build time reduction, shortening product release cycles from multi-month to 4-6 weeks, and migration to ubiquitous GitHub to further optimize the organization's global development capability.
‘Analytics as a Service’ telemetry pipeline
Challenge
Provide machine data analysis for a population of over 100M devices in a service provider network. Predict network failures and identify performance issues that could impact customer satisfaction and mitigate accordingly. Provide a data lake of relational telemetry data that could be consumed by engineering, business development, and operational teams across the organization. Aggressively seek to minimize operational costs.
Solution
Vypian staff managed the team that developed and deployed an AWS cloud based telemetry system including S3 storage, EC2 compute, and supporting components. Collecting runtime logs and JSON telemetry data from over 100M devices totaling 30TB per day. Employed elastic scaling to optimize resource usage and minimize costs. Provided a data lake of relational telemetry that could be directly consumed by business stakeholders and allowed analysts to exercise their own algorithms.
Result
The system allowed operations teams to identify potential customer-impacting anomalies and address them before becoming a problem, significantly improving customer satisfaction. Consumers of the system's curated data extended beyond traditional engineering/operations teams to include business development, product management, and marketing teams. Continuous system optimization and data consolidation resulted in >$1M OPEX reduction. Furthermore, the coherent system design made it possible to rapidly deploy new telemetry pipelines in greenfield locations around the globe.
AI creation of software test suite
Challenge
A test suite was required to exercise approximately 60 low-level software components (HALs) that are delivered from multiple 3rd party vendors and form part of a larger embedded software stack. The specification for each HAL is well-defined and describes how the component is expected to integrate and operate with other software components in the stack. The fundamental idea behind the test suite is to ensure any HAL components are 'perfect' upon initial delivery, avoiding more expensive problems being discovered later in the development cycle. To enable ease of use, the tests had to be consistent in approach and operation, intuitive, and have minimal dependency on external infrastructure including physical test harnesses or the need for a full software stack before testing could commence. Test suite installation had to be simple and low maintenance; the output from the tests needed to be reported in a concise manner to aid rapid assessment of a deliverable's quality.
​
Solution
An AI Large Language Model (LLM) was employed to create individual test cases using the following steps:
-
The HAL documentation was refined to ensure it followed a standard industry format (Doxygen, https://www.doxygen.nl ). All documentation was captured in the HAL source code header files.
-
Suitable AI prompts were created that instructed the LLM to parse the header file to understand the HAL’s function and expected operation, then generate appropriate test cases that could exercise the HAL's behavior.
-
LLM generated tests were reviewed by HAL owners/architects for accuracy and adjustment.
-
Where necessary, the LLM was iteratively asked to adjust the test code until production quality was attained. Various AI techniques were employed to ensure consistency of the test cases including ‘retrieval augmented generation’ (RAG), fine tuning, and ‘prompt baking'.
Result
Over 3000 test cases were created using AI that comprehensively exercised the 60 HAL interfaces. Quality of software deliverables increased significantly leading to down-stream cost reduction. AI allowed the test suite to be delivered in the order of weeks; without AI it would have taken several engineer-years to reach the same outcome. The resulting test suites are in operation with multiple software vendors.
AI facilitated database migration
Challenge
A music catalog stored in a legacy database (30 years in operation) was required to be migrated to a MySQL database. The original database contained information for over 7000 entries spanning 35 unique fields. The MySQL database was to use a new design that would resolve any architectural weaknesses present in the legacy design, including inconsistent data formats (e.g. multiple time/date formats scattered throughout the database), and suboptimal data grouping that made query creation over-complicated or the results too granular to be useful.
Solution
Vypian managed the project which took the following approach:
-
Perform a thorough analysis of the legacy database including the architecture and data set. Identify problems and weaknesses.
-
Document the requirements of the new database. Where still relevant, combine original legacy requirements with new requirements; discard any legacy requirements that were no longer applicable.
-
Using MySQL Workbench, design the new MySQL database architecture using Enhanced Entity Relationship (EER) diagrams to enforce a valid and logical design.
-
Validate the architecture against the requirements. The new design would require over 40,000 entries spanning over 50 unique fields.
-
AI was extensively used to during the data migration phase to clean-up the legacy data (for example, to identify dates and apply a consistent format to all instances) and perform the mapping of the data from the legacy 35 fields to the new 50+ fields.
Result
The legacy database was successfully migrated to a MySQL database. AI was essential to the success of this project given the size of the migrated data set, the need to clean-up the content before migration, and the number of field transformations that were required. The migration took a few weeks to complete; without AI is would have taken several engineer-years to complete and a significant amount of scripting.
Vypian breathed new life into a database that would otherwise have been abandoned. Really appreciate the work!
Primaudial Band
October 2024
Adoption of AI
Challenge
With the industry surge of interest in the promise of AI, a large, global engineering organization in a Fortune 100 company was tasked with exploring the potential benefits of AI and utilizing it to solve real business issues. To ensure any benefits of AI could be easily measured, regular business operations that involved significant overhead or were tedious by nature were chosen as case studies to validate the approach. Examples of such projects included data retrieval from various corporate information systems, analysis of logs/telemetry from products in the field, avoidance of data duplication, and helping software developers write code faster. At all times during this adoption period, relevant IP protection, privacy, and copyright laws had to be observed.
Solution
Vypian staff created and managed a new team of carefully selected engineers to work on various AI approaches. This team was responsible for all aspects of the work streams including:
-
Vetting the many ideas solicited from the organization and choosing those to pursue based upon potential benefit to the business.
-
Working with the corporate legal team to approve all aspects of a case study before proceeding with engineering work. This included approving the AI model to use, approving the data set that would be fed to the AI model, and approving the use of any output from the model.
-
Developing the AI engineering prompts and procedures.
-
Developing governance that would be used across the organization as AI use expanded beyond the initial ‘Skunk Works’ team.
Result
The team established numerous AI large language models, either running on isolated GPUs or through managed services such as Azure. These models were used to run many 'proof of concept' workstreams, with those demonstrating real benefit to the business being productized and rolled-out across the organization.
Examples of projects that were productized and deployed included an organization-centric ‘Ask Me Anything’ portal and a utility that interfaced with the organization’s bug tracking system to prevent duplicate entries from being created. Both products were widely used and helped build acceptance of AI across the organization.
​
Some of the case studies proved more challenging than expected, including one designed to analyze masses of telemetry data to identify deep-rooted software bugs. This effort did show promise, but the disparate nature of the telemetry data made it difficult for the AI to produce consistent, meaningful results. However, these tougher problems provided opportunities to learn about AI behavior and helped accelerate its use across the group. As a result of this early adoption effort, AI is now being used as a standard tool across the organization.
I am delighted to extend my highest recommendation for Gary Jamieson and his exceptional leadership at Vypian. Gary has been a pivotal figurehead in spearheading the development and implementation of in-house AI proof of concepts, tools, and services across our global development organization, particularly in our Fortune 500 company. His deep-rooted relationships throughout the company facilitated meaningful engagement and secured substantial commitments from senior leaders in terms of both time and financial investment. The successes achieved through these AI-driven initiatives have been transformative, resulting in thousands of lines of AI-generated code, enhanced product testing, superior product quality, and notable improvements in cycle times and resolutions. Moreover, Gary’s efforts have significantly elevated AI awareness within the company, setting a strong foundation for continued innovation and growth.
AI Evangelist
November 2024