Full Stack Engineer
Designed and built the company's production platform (six services on AWS, the database behind it and an on-device vision pipeline) and led the team's two interns after the lead developer left.
- —Replaced manual deployments with the company's first containerized platform: six services on AWS EC2 (two FastAPI microservices, Label Studio, MLflow, Nginx) brought up reproducibly with Docker Compose.
- —Built an on-device computer-vision pipeline on NVIDIA Jetson Orin Nano that identifies the products a customer takes from a retail cooler, plus the evaluation harness that set its production requirements.
- —Designed the PostgreSQL schema the operation runs on: 25 tables on AWS RDS with 24 foreign keys, 34 indexes, idempotent migrations and a soft-delete audit trail, reachable only through an SSH bastion.
Show 4 more
Show less
- —Built an event-driven FastAPI webhook service between Label Studio and S3 that keeps the two consistent when a step fails: atomic rollback across database and S3, async connection pooling, a cache with database fallback. Wrote the runbooks to operate it.
- —Moved the frame-review desktop app off Windows-only WPF to cross-platform C# (.NET 10, Avalonia): MVVM, two-pass verification and a sliding-window prefetch cache.
- —Locked down access with least-privilege IAM, AWS Secrets Manager and LocalStack, so development never touches production credentials; set up the CI pipelines.
- —Built the data pipeline's operator tooling (CLIs, idempotent S3-to-PostgreSQL sync, drift audits) and automated model training in GitLab CI on ephemeral AWS GPU instances, tracked in MLflow.
FastAPIDocker ComposeAWS EC2 / RDS / S3PostgreSQLNginxNVIDIA JetsonTensorRTC# · AvaloniaGitLab CI