Loading...
Loading...
A showcase of professional work and personal projects exploring platform engineering, MLOps, and cloud-native technologies.
In-depth looks at platform challenges, approaches, and outcomes.
Data-intensive analytics company processing intelligence data across multiple regions, requiring a modern data platform to handle 50TB+ daily data with strict governance requirements.
The existing data infrastructure couldn't scale to meet growing data volumes. Manual cluster management led to inefficient resource utilization, and lack of standardized pipelines created inconsistency across data teams. Cost visibility was poor, making optimization difficult.
Designed a multi-region Databricks platform with Infrastructure as Code at its core, implementing medallion architecture for data organization and self-service capabilities for data teams.
Global fashion retailer (Calvin Klein, Tommy Hilfiger) running 200+ microservices on legacy VM infrastructure, facing scaling challenges and slow deployment velocity.
Weekly deployments were the norm, with each taking 2+ hours. Teams waited days for infrastructure provisioning. Observability was fragmented across tools, making incident response slow. The VM-based infrastructure couldn't efficiently handle traffic spikes.
Built a Kubernetes-native platform with self-service capabilities, comprehensive observability, and GitOps-driven deployments. Migrated services incrementally with feature flags to minimize risk.
Large-scale observability requirements for 200+ services generating terabytes of logs daily, with existing Elasticsearch clusters becoming increasingly expensive and difficult to manage.
Elasticsearch licensing costs were escalating rapidly. Cluster management was manual and error-prone. Index lifecycle management was inconsistent, leading to storage bloat. Teams lacked self-service capabilities for creating dashboards and alerts.
Migrated from the ELK stack to AWS Managed OpenSearch with a focus on cost reduction and operational efficiency, implementing automated index lifecycle management and self-service patterns for development teams.
Healthcare SaaS needing a secure, EU-compliant platform for an LLM-powered clinical trial matching API. B2B API product processing de-identified medical records through multi-model LLM inference.
No existing infrastructure for the new Matching API product. Needed dedicated isolation for medical data compliance, LLM inference co-located for zero-latency, SQS-driven async processing with autoscaling workers, and defence-in-depth security for a regulated healthcare environment.
Designed and delivered a dedicated EKS cluster with defence-in-depth security, co-located LLM proxy, CloudFront VPC Origins for zero-exposure production APIs, and KEDA-driven autoscaling for async batch LLM processing.
Multi-cloud MLOps reference architecture enabling self-service model deployment across AWS EKS, Azure AKS, and GCP GKE with defense-in-depth security and full-stack observability.
The problem this architecture addresses: ML teams commonly lose 2-3 days per model deployment to manual processes, ad-hoc security reviews, and invisible GPU spend — with no consistent path from experimentation to production across clouds.
Built a multi-cloud MLOps reference platform with security-first infrastructure, self-service model serving, and per-cloud Terraform modules across three providers.
Personal projects exploring MLOps, AI infrastructure, and advanced Kubernetes patterns.
Kubernetes-native AI inference gateway built from scratch to explore the multi-model routing problem space: A/B testing, intelligent failover, circuit breakers with exponential backoff, OpenTelemetry tracing, and cost/latency/context-length-aware routing. The ecosystem has since standardized this layer (Gateway API Inference Extension, llm-d) — the learnings from building it independently now feed my upstream contributions there.
Reference implementation of a cost-optimization platform for AI/ML workloads: GPU utilization monitoring, budget forecasting with alerts, ML-based anomaly detection, automated right-sizing recommendations, and multi-cloud billing integration. Code-complete across all 3 phases; validation against live GPU clusters is the next milestone.
Multi-cloud MLOps reference platform on AWS EKS, Azure AKS, and GCP GKE with defense-in-depth security and full-stack observability. Designed around a 15-minute self-service path from experiment to production; end-to-end verification, LLM-serving benchmarks on real GPUs, and GitOps rollout are the current focus.