PresentContractStaff since Oct 2024 · Senior Software Engineer, Feb–Oct 2024

Staff Software Engineer
Software Sushi · Miami, FL (Remote)
- Designed MAGGIE end to end — ThreatMinder's AI entity-intelligence product for risk, fraud and identity vetting: domain model, API, data stores, agent layer, batch pipeline, back-office control plane and billing, with 12M+ documents processed
- System design across four stores: a NestJS API on PostgreSQL, DynamoDB for entity searches and usage, Redis for cache and metering, and OpenSearch for retrieval — on ECS with Lambda, Kinesis and S3, including the pagination, indexing and graceful-degradation rules every other service follows
- Author of @maggie/core, the internal library every service imports: typed entity-search models, DynamoDB repositories, a traced Envelope message contract and centralized logging, published to CodeArtifact on merge
- Made model behaviour governable — a versioned use-case library (prompt, model and risk categories per version, gated rollout) so scoring changes ship as reviewable versions rather than code edits, alongside LangGraph workflows and MCP endpoints shipped as product surfaces
- Designed the metering and billing layer: usage and AI-cost events with period aggregates, a cost-rate catalog and per-organization spend limits — the platform prices, bills and throttles itself, and degrades instead of failing user requests
3 more from this role
- Rebuilt the cloud underneath it at ThreatMinder: the monthly AWS bill ~86% lower (~$25k → under $3.5k) while the platform got faster; manual zip deploys replaced by infrastructure-as-code and CI/CD; Postgres CPU from a 58% average (peaks near 100%) to 0.3%, then 48 → 8 vCPUs
- Sole architect of the Insurate document-AI pipeline: model bake-off → FastAPI → an 82-state Step Functions workflow, 16 Lambdas consolidated to 8, 97.7–100% field match against the incumbent vendor on a 179-PDF eval corpus
- Grew the practice — took a small consultancy from internal bots to paid ML pipelines and production AI agents: scoping engagements, setting the delivery bar (IaC, CI/CD, observability, evals), mentoring engineers, and owning SEV response and on-call
Go · Python · AWS · GCP · Kubernetes · Terraform · Packer · LLMs · Kafka · PostgreSQL



