Adam Drąg

Australia, remote

Engineering calculation platform: reactive MDX documents with live, dependency-graph calculation blocks over MathJS, Python and spreadsheets.

• Built the platform’s agentic AI assistant end to end on AWS Bedrock: 13 tools over a 14-step agentic loop, a validation repair cycle and prompt caching, with user-selectable models (Claude, Kimi, GLM).
• As the team’s primary code reviewer, introduced AI-native development workflows with Claude Code: custom testing and domain-knowledge skills, plus automated PR review.
• Cut AWS spend 25% by decommissioning idle resources, retuning ECS Fargate autoscaling, right-sizing tasks against 30-day CloudWatch data, and migrating to Aurora Serverless PostgreSQL.
• Engineered the dataset pipeline: event-driven CSV → Apache Parquet conversion on S3 with DuckDB, plus Bedrock knowledge-base ingestion powering RAG over workspace documents.
• Shipped user-facing usage-based billing that monetises the assistant: per-model per-token pricing across 5 models and 3 providers, Stripe metering, entitlements, trials, org budgets and per-event rate snapshots.
• Migrated the production fleet to ARM64 Graviton Fargate (~20% cheaper compute), with multi-arch amd64/arm64 Docker images built in GitHub Actions for zero-downtime rolling deploys.
• Owned observability and incident response: wired Datadog RUM/APM with source maps, diagnosed a webhook pile-up that OOM-killed the collaboration service, and cut prod log volume 97% on a 19.7 GB log group.
• Built the local-first React 19 frontend: an MDX rich-text editor (Plate.js) with math and spreadsheet blocks, Yjs CRDT collaboration and cursor-based GraphQL sync into IndexedDB.
• Architected the calculation engine: a DAG over MathJS, sandboxed Python on AWS Lambda and Parquet datasets, recomputing only the statements downstream of a change.
• Ported the platform to a Tauri desktop app, integrating over COM with Microsoft Excel and ETABS to read and write live engineering models.

Australia, remote

• Built the automated billing and provisioning platform for a cloud-native vector database, deploying per-customer instances on AWS OpenSearch.
• Engineered a usage-metering pipeline aggregating AWS instance metrics every 15 minutes, wired into Stripe to automate usage-based invoicing.
• Automated provisioning and scaling of the managed vector-search offering, removing manual steps from customer onboarding.

Australia, remote

• Built backend APIs and reporting services for the logistics performance team at Appliances Online, one of Australia’s largest e-commerce retailers, tracking delivery KPIs nationwide.
• Maintained and optimised data-aggregation systems giving operations visibility into nationwide delivery and fleet performance.

Wrocław, Poland

• Principal developer on large-scale enterprise systems, owning architecture, framework selection and design across multi-year client engagements.
• Led and mentored cross-functional teams of 2-4 engineers, holding the team to Clean Code and software-design standards.
• Built backend services in TypeScript and Java, and frontends in Angular, React and Vue.js.

Wrocław, Poland

• Estimated, designed and delivered client and server-side banking applications in an Agile Scrum team, maintaining quality through peer review.

Wrocław, Poland

• Full-stack developer on Nokia’s private cloud platform: provisioned VMs, designed the Java and Node.js microservices running on them, and built the web frontends.
• Created and managed CI/CD pipelines, ran code reviews and worked in Agile Scrum planning.

Oleśnica, Poland

• Built standalone Java applications and Excel VBA automation that cut manual work in material call-off and production planning, and controlled stock, purchasing and delivery planning.