Fifteen years of cloud-native platforms and applied AI — across finance and healthcare. Underneath the domains it's one problem: finding the structure that lets something complicated stay correct, fast, and easy to change.
Most systems don't fail on the happy path. They fail at the seams — the integration nobody owned, the resource left running, the assumption that held until it didn't. The job is to find those seams before production does.
Get the data model and the boundaries right and the code mostly writes itself. Get them wrong and no amount of clever code saves it. The expensive thinking goes up front, where it's still cheap to change.
What a system reports and what it actually did are never quite the same system. Design for observability and check the running reality — the metric, the trace, the behavior — not the deploy that “succeeded.”
Resilience isn't heroics. It's circuit breakers, idempotency, graceful degradation, and treating an outage as a tuned-for event rather than a surprise. The best systems fail quietly and recover on their own.
Some client names are confidential and are described by sector and scale. The architecture is the point either way.
Worked for an American consumer credit-reporting agency, integrating data from many disparate sources into a single platform that pairs AI/ML algorithms with human governance to produce accurate credit-risk reports for consumers. Designed the ingestion pipelines, the microservices that process them, and the model and review workflows that keep the scoring both accurate and accountable.
Built the ML layer on AWS SageMaker — training, deployment, and monitoring of the risk models — with S3 and Redshift feeding the feature data, and KMS, IAM, VPC, and WAF carrying the security model for regulated financial data.
Architected a multi-layer healthcare platform — React front end, Java/Spring and Python/Django services, and an AI/ML layer in TensorFlow and PyTorch — to process patient records, device telemetry, imaging, and genomic data. Built a medical-intelligence agent on AWS using foundation models for clinical decision support.
Designed the whole thing to be compliant by construction: AES-256 at rest, TLS in transit, a Hyperledger Fabric ledger for auditability, and ingestion pipelines built for HIPAA and GDPR from the first line.
Modernized a legacy enterprise tool into microservices on RedHat OpenShift — standing up an Istio service mesh with Envoy sidecars for traffic management and inter-service security, OIDC/OAuth 2.0 for identity, and a Kafka-fed audit-logging system writing real-time event trails into Db2.
Brought the platform a full observability stack — Kiali, Prometheus, Grafana — so service health and performance were visible rather than inferred.
Led the migration of a personal-investment pre-clearance system — UI, API, and database — onto AWS, collaborating with business users on the roadmap and migration plan. Moved services to EKS, the database to Aurora MySQL, and wired up Okta auth, Kafka, and Splunk for the new environment.
Built the CI/CD path on GitHub Actions with automated deployment to EKS, and moved analytics reporting from on-prem to AWS-hosted with secure bridge connectivity.
Architected REST APIs and a self-service PaaS portal on AWS EKS, integrating Jira, Confluence, GitHub, and Artifactory. On a parallel program, designed an event-driven aircraft-maintenance system using the Axon framework and CQRS, with Kafka and RabbitMQ tuned for durability and fault tolerance.
Applied Resilience4j circuit-breaker and bulkhead patterns to stop cascading failures, and tuned services to cut CPU overhead and prevent pod restarts under load.
What I actually reach for, grouped by where I spend my days — not by how many keywords I can list.
Before software there was physics, and before the job there was a long fascination with how things actually work underneath — the parts you can't see that decide whether the visible thing stands or falls.
Physics teaches you to distrust the surface and look for the law underneath it. Biology teaches you that complicated systems are usually simpler than they look once you find the right level to look at them. Somewhere between those two habits is most of what I do now.
So a banking platform, a clinical model, and a piece of cloud infrastructure stop feeling like different jobs. They're the same quiet question asked in different clothes: what's really going on here, where does it want to break, and what's the smallest structure that keeps it honest. I'm at my best where something is genuinely tangled and someone needs to find the shape that makes it make sense again.