I've spent years in the weeds of cloud-native infrastructure — managing VM fleets, debugging cold starts, architecting distributed job schedulers, and helping developers ship AI applications faster. This is where my mathematical background and hands-on engineering instincts converge.
Architected and shipped a serverless platform enabling GPU-based AI inference, web scraping, and video processing at scale. Managed GKE-based infrastructure, production CI/CD pipelines, and multi-vendor cloud provisioning from day one — automating change management to ensure seamless deployments across heterogeneous environments.
Visit ByteNiteDirectly supported developer customers deploying open-source LLMs, diffusers models, and data pipelines on the platform. Diagnosed GPU cold start issues, instrumented model quality checks, and delivered targeted solutions that enabled five production deployments within three months, while also cutting average setup time by 3×. The work required both systems-level debugging and clear communication with non-infrastructure teams.
Invented and patented a system for dynamic partitioning and orchestration of computational jobs across heterogeneous compute grids, optimizing for capacity and availability. Patent granted in both the United States and European Union. The core insight: treating compute supply and demand as a scheduling problem solvable with resource-aware algorithms.
View patentAt ByteNite, I managed GCP production infrastructure daily — Compute Engine, GKE, and Cloud Storage — staying close to the cost and performance trade-offs of a multi-tenant compute environment. To round out that experience, I completed an AWS Solutions Architect certification covering architecture design, networking, security, and CDN configuration at the AWS service level.