US enterprise software vendor whose analytics turn performance and cost data into infrastructure decisions for hardware and cloud buyers. Responsible for the AWS platform behind those products across its whole lifecycle, from architecture and delivery to production operations.
Cloud & Platform Engineer
Sep 2025 – Present- Own and run the serverless recommendation platform behind the public advisory tool a major silicon vendor offers its customers: infrastructure as code with AWS CDK in Python (API Gateway, Lambda, DynamoDB, S3, EventBridge), one stack definition for every stage from sandbox to production, deployed stage by stage with post‑deploy Synthetics canaries and alarms as the release gate.
- Designed the analyst delivery path for the recommendation models: a notebook dropped in S3 is validated, compiled into the API handler and promoted through the same stages and canary checks as any release, with rollback to the previous handler on failure, so algorithm changes reach production without hand‑written application code. The pattern now runs multiple production APIs.
- Ran the performance side of a price‑performance program: SPEC CPU2017, STREAM and FIO across hundreds of EC2 instance types in every current family, on a self‑terminating cloud‑init fleet, within a fixed budget, with pilot campaigns on GCP and Azure. Results joined with the pricing dataset for cost‑per‑performance comparisons and published as branded reports behind CloudFront.
- Shipped LLM features to production: Amazon Bedrock schema mapping inside a Step Functions Distributed Map for customer data intake, a Claude personalization endpoint on recommendation results, a Claude vision pipeline with tool‑forced structured output, and an asynchronous LLM job pattern on SQS and DynamoDB.
- Built the product analytics pipeline on Step Functions and Lambda: daily usage telemetry validated and aggregated into partitioned Parquet on Athena and Apache Iceberg, with MERGE upserts for month‑end restatement, SNS reporting and a one‑command backfill.
- Set the operating conventions for every CDK stack and enforced them: TTL on transient DynamoDB tables, lifecycle rules on working buckets, retention on every log group, shared dependencies in Lambda layers rebuilt per runtime, and a canary‑and‑alarm pattern for each new API.
- Led an estate‑wide secrets audit across every repository and deployed Lambda function and moved every service off long‑lived keys onto IAM roles and Secrets Manager.
- Handled production incidents end to end and turned each into a control: an empty upstream pricing feed became an abort gate on the weekly refresh; a rollback that fired on every failure became per‑query deadlines and stage‑level alerts on the analytics pipeline; a dead‑letter handler's first poison message became idempotent processing with an explicit retry policy.
Full Stack AWS Cloud Engineer
Sep 2022 – Sep 2025- Re‑architected the customer data processing engine, which turns spreadsheet exports from OEM and enterprise customers into normalized datasets, from a Lambda pipeline into a containerized Java 17 / Spring Boot 3 service on ECS Fargate: high‑memory tasks, S3‑event triggers, DynamoDB status tracking, ECR image rollouts and a declarative set of transformation actions, removing the Lambda time and memory ceilings.
- Engineered the large‑dataset processing layer of that engine with Apache POI: streaming parsers, JVM memory management (ZGC), cross‑file joins and a custom formula engine for inputs of several hundred megabytes, validated against a regression suite of real customer data.
- Developed the multi‑cloud pricing dataset behind the cost models: AWS, Azure and GCP price lists collected, normalized to one schema and filtered to comparable SKUs, refreshed weekly with change detection, and served through Athena for price‑performance and TCO comparisons.
- Maintained the product's front‑end runtime (Webpack, Chart.js 4) and built its embeddable LLM chatbot widget with async submit‑and‑poll and file upload.
- Built the visual‑regression QA service that runs before every product release: Puppeteer screenshots on ECS Fargate, pixel diffs and an annotated Excel report; and the headless‑Chrome print service that replaced a legacy ESB export path for customer‑facing PDFs.
- Inventoried the Lambda estate and ran the Python runtime cleanup: triaged every function, retired the dead ones with archive‑and‑redeploy tooling, and upgraded the rest to Python 3.12.
Backend & Cloud Engineer
Apr 2021 – Sep 2022- Built the first serverless generation of the customer data processing engine on AWS Lambda (Node 18, ExcelJS, HyperFormula): validation, action execution, S3‑event architecture and DynamoDB status tracking.
- Built the customer benchmark intake pipeline, portal upload to S3 to JSON to Athena tables and views with SES notifications, still running in production five years later.
- Owned deployments end to end: Lambda packaging, IAM roles, S3 event wiring and DynamoDB data modelling across Node.js and Python services.
Earlier roles
Raspberry Pi based remote pet‑monitoring and live‑streaming system: cellular connectivity, RTMP/HLS video, a Firebase and AWS backend and a Flutter app.
Second engagement with the same client: rebuilt the industrial cleaning‑robot control app in Flutter for the startup's seed‑stage product, replacing the 2019 native Android proof of concept; robot control over MQTT.
Companion app for a family of Wi‑Fi sensor devices: onboarding over the device's own access point, alerts and push notifications, per‑sensor thresholds and charts, on an AWS Amplify backend (Cognito, Lambda).
Spare‑parts catalogue for a Canadian all‑terrain‑vehicle dealer: exploded‑view drawings with clickable image maps, revision‑aware part tables and e‑mail ordering, for three vehicle models.
Android application for Fourier's einstein Tablet+ classroom science tablets: physics and chemistry lessons as live experiments, driven by built‑in sensors, external lab probes and Bluetooth sensors through the vendor SDK, with real‑time charts and a USB camera feed; three languages, three tablet generations.
Proof‑of‑concept control app for a fleet of industrial cleaning robots over MQTT.