Case Studies

Challenge

A major government services organization operating one of the largest contact center environments in its sector needed to migrate its entire telephony infrastructure to the cloud. The environment handled nearly 2 million calls per month across 50+ regional offices, each with unique operational requirements, legacy integrations, and partner dependencies. No repeatable migration framework existed, and the organization had limited cloud experience.

Approach

• Designed and led a multi-wave migration strategy spanning six deployment phases, coordinating across customer leadership, systems integration partners, and internal cloud architecture teams
• Built an on-site support model assembling cross-functional teams of solutions architects, customer success managers, and sales specialists for each go-live event
• Developed a go-live communication plan with real-time escalation paths reaching senior executive leadership, ensuring full visibility into risks, blockers, and customer sentiment
• Created a reusable wiki, business model map, and onboarding documentation to support the dozens of resources rotating through the engagement
• Collected and incorporated lessons learned from each migration wave to continuously refine the process and reduce deployment risk

Impact

Delivered the largest cloud contact center deployment in the region. The go-live support model was presented as a best practice to 60+ customer success managers and adopted as a standard methodology. Identified and positioned $3.5M in additional service adoption opportunities post-migration, including voice biometrics, global resiliency, and AI-powered language support. Bridged CSM leadership coverage for a $25M ARR account representing 20% of the regional portfolio.

Challenge

A leading education technology platform serving over 35 million users across thousands of institutions needed to build an intelligent AI assistant capable of automating complex workflows across its ecosystem. The platform exposed 700+ APIs, and the assistant needed to orchestrate actions across them while maintaining strict quality control, security standards, and a feedback loop for continuous improvement. No prior AI agent architecture existed within the organization.

Approach

• Led cross-functional product and technical discussions to scope the AI agent's capabilities, defining the integration strategy across 700+ platform APIs
• Designed an intelligent workflow automation system with metered access controls and a quality assurance framework to ensure reliability at scale
• Developed a user feedback mechanism enabling educators to rate and refine AI-generated actions, creating a human-in-the-loop improvement cycle
• Coordinated alignment across the customer's CTO, CPO, Chief Architect, and Head of AI to ensure executive sponsorship and technical buy-in
• Structured the rollout in phases, starting with high-impact administrative tasks to demonstrate value before expanding to more complex workflows

Impact

Delivered a production AI agent that simplified complex administrative tasks for educators while maintaining platform security and performance standards. The implementation became a flagship AI initiative for the customer, driving increased user satisfaction and improved retention rates. Positioned the cloud provider as a strategic AI partner for the customer's broader transformation roadmap, unlocking additional AI/ML workloads across the organization.

Challenge

An enterprise education platform was preparing to launch a skills alignment product connecting academic coursework to workforce outcomes. The core blocker was an AI-generated skills taxonomy that required 37% manual correction, making it impossible to scale ahead of the planned general availability launch. The platform needed a production-ready solution that could accurately extract skills from diverse course content and map them to standardized taxonomies with minimal human intervention.

Approach

• Managed a 6-week engagement with a dedicated AI innovation team, coordinating between cloud engineers, product managers, and customer subject matter experts
• Designed a three-step pipeline architecture: module-level content processing, LLM-as-judge evaluation methodology, and human-in-the-loop validation
• Implemented taxonomy seed functionality enabling customer-specific customization, learning outcomes integration, and multi-course skills coordination using vector stores
• Built comprehensive feedback mechanisms that allowed the system to learn from corrections and progressively reduce manual intervention rates
• Delivered full code documentation and production deployment guides to ensure the customer could independently operate and iterate on the solution

Impact

Removed the primary technical barrier preventing the customer's product launch. The solution processed 11 courses from 5 different customer institutions with near-perfect precision and recall scores after incorporating feedback mechanisms. Established a reusable architecture pattern for educational skills extraction that can be applied across the customer's 30+ million global user base. The engagement directly enabled a $300K ARR product launch and strengthened the strategic partnership with a $61M ARR account.

Challenge

A major K-12 education technology platform needed a content moderation solution for its AI-powered student assistant. The existing implementation on a competing cloud provider suffered from unacceptable false positive and negative rates, failing to meet the strict safety standards required for environments serving millions of children. Off-the-shelf AI safety guardrails were not yet mature enough to handle the nuances of educational content, and the platform needed a custom solution that could be deployed to production without compromising student safety.

Approach

• Led cross-functional coordination between cloud account teams, ML service specialists, and the customer's offshore development teams to define the solution architecture
• Designed and oversaw implementation of a custom ML model for hateful and abusive language detection, trained specifically for educational contexts using cloud-native ML infrastructure
• Established a validation framework for production deployment, ensuring the model met strict safety thresholds before rollout to student-facing environments
• Created a bridge solution that maintained safety standards while longer-term platform guardrails matured, preventing any gap in content protection
• Worked directly with the customer's senior AI leadership to align on requirements, testing protocols, and deployment timelines

Impact

Delivered a production content moderation system that significantly outperformed the previous competitor implementation on false positive and negative rates. The solution provided continuous protection for over 3 million K-12 students while the platform's broader AI safety capabilities matured. Strengthened the strategic relationship with a major education technology account and created a reusable framework for custom content moderation that can be applied across similar regulated environments. The engagement also generated direct feedback to cloud AI safety product teams, contributing to the improvement of platform-native guardrail features.