Selected Programs & Technical Work

Situation

An enterprise voice platform operating at 1B+ sessions per month and $250M+ revenue scale, built on legacy RNN models — in the midst of an acquisition, serving Fortune 500 customers with strict continuity requirements.

Task

Modernize the platform to LLM-driven, agent-ready architecture while leading the organization through acquisition integration and sustaining delivery, with no reliability regression.

Action

Defined and executed the AI modernization roadmap with product leadership toward an agentic-AI and copilot strategy. Rebuilt the model experimentation pipeline into a continuous release cadence. Migrated self-hosted inference to managed cloud AI services where economics warranted. Established AI governance and secure engineering practices — DevSecOps, secure SDLC, shift-left tooling — as the platform baseline.

Result

RNN to LLM real-time inference at <100ms latency. Experimentation cycles reduced from multi-quarter to 2 weeks. Inference costs reduced ~30%. Reliability improved 25% and incident resolution time reduced ~40% at 99.9%+ availability. 90% of senior staff retained through the first three years of integration.

Situation

Monolithic and VMware-era systems serving regulated customers under PCI-DSS, HIPAA, SOC 2, and SOX — customers for whom extended migration windows were not acceptable — on infrastructure constraining platform growth.

Task

Migrate the estate to cloud-native microservices across global Azure regions within nine months, with zero compliance regression.

Action

Designed the migration architecture around the continuity constraint: tenant-by-tenant migration with tenant-by-tenant rollback, eliminating dependence on large cutover events. Chaired the Architecture Review Board through each consequential decision, and governed build-versus-buy at the platform level.

Result

10,000+ nodes migrated in 9 months with compliance posture intact, on Azure and AWS — establishing the foundation for platform revenue growth from $15M to $250M+.

Situation

Forty data centers of accumulated footprint, with data-residency obligations across multiple jurisdictions and an infrastructure cost curve outpacing business growth.

Task

Consolidate the footprint and reduce the cost base without compromising residency commitments to regulated customers in Canada, Australia, the UK, EMEA, and the UAE.

Action

Established data residency as the primary design driver, with cost as the measured consequence. Built global edge and point-of-presence infrastructure through Equinix and cloud-exchange providers, decoupling regional presence from regional data centers. Directed the $40M+ infrastructure and vendor budget through the transition.

Result

40 data centers consolidated; approximately $96M (~40%) removed from annual infrastructure cost while expanding regional coverage.

Situation

A global voice backbone carrying 200,000+ concurrent sessions across international points of presence, degrading under load during a major financial-services engagement — in a real-time medium where degradation is immediately customer-visible.

Task

Stabilize performance and scale for the engagement while protecting service for all tenants on shared infrastructure.

Action

Re-architected platform distribution across points of presence and resolved the scaling behavior at the platform level rather than as a single-customer remediation, so the correction generalized across the tenant base.

Result

The engagement stabilized, and the platform baseline was elevated to federal standard — extending the improvement to every customer on the backbone.

Situation

A company at inception — prior to product or additional staff — requiring authentication, authorization, and accounting infrastructure before subscribers could be provisioned or billed.

Task

Design and build the authentication and access infrastructure on which the business would operate.

Action

Architected the AAA/RADIUS authentication layer and constructed the surrounding infrastructure, scaling it in step with subscriber growth.

Result

Production infrastructure serving thousands of subscribers — a zero-to-one build that established identity and access management as a career-long specialization.