resume

Download PDF · LinkedIn · GitHub · john@connol.ly

Principal Engineer · AI Engineer: LLM apps, RAG, evals, agents, MCP · Product & Forward-Deployed Engineering · Python, TypeScript, Azure, AWS, Terraform

experience

Data Society Group · Lead Product Engineer

Principal-level IC · Remote · Jan 2025 – Present

Ran engineering, delivery and product for a construction client’s program while building the backend and running the infrastructure myself.

Project intake and financial operations platform (bid to accounting and operations, ERP integration)

• Reviewed the proposal and estimate early; shaped the schedule, roles and split of work across onshore, offshore and partner developers.
• Solutions architecture: Figma designs to data model and OpenAPI spec. Designed and built the integration with the client’s ERP: contract and project creation, approval callbacks, outbox dispatch, reconciliation.
• Led delivery to go-live with the client and integration partner: Jira program, sprint planning, stakeholder reporting.
• Playwright end-to-end tests confirming records landed in the ERP.
• AI-assisted development: parallel Claude Code agents in isolated devcontainers and git worktrees, adversarial PR review, verify-before-merge CI gates. Documented it for the team.

Generative AI document Q&A pilot on Azure (Jun 2026 – present)

• Ingestion: Azure AI Document Intelligence and native Office extractors produce validated JSON, chunked and embedded into pgvector. Hybrid retrieval with version-aware answers.
• Golden-question eval harness with a recorded baseline, turn logging and a conversation audit trail. An end-to-end pipeline integration test caught a real cross-PR bug.
• Azure infra in Terraform: Container Apps and Jobs, GitHub OIDC deploys, Entra ID JWT auth. The agent UI runs as a Module Federation remote inside the client’s app.

Internal / platform

• Stood up the Post-Course Reporting app (Next.js App Router, Tailwind 4, Prisma, ECS Fargate behind ALB OIDC + Google Workspace SSO), fed by a Glue/Python job mirroring a partner’s Redshift schema into RDS Postgres.
• Owned Terraform (ECS Fargate module, Route 53, ECR, RDS, cross-account IAM) and Bitbucket Pipelines (build → ECR → terraform apply → ECS/Lambda) for the AI/platform side.
• LLM teaching assistant (Amazon Bedrock + Claude, RAG over course content) and AI course recommender (embeddings + vector search) on AWS ECS/Lambda.
• dsg-tools: an MCP server and Claude skills the engineering team uses daily for Jira, Bitbucket and AWS.

Career break

Feb 2024 – Dec 2024

Planned break for family time and hands-on work. Restoring a 13’ Boston Whaler (stripped to bare fiberglass; sanding, fairing, gelcoat; mahogany trim to come). Worked on a 1993 Volvo to learn car mechanics from the ground up.

Flow Commerce · Principal Software Engineer

Acquired by Global-e (NASDAQ: GLBE) · Jul 2021 – Feb 2024

• Tech lead for 3–5 engineers on the checkout, payments, tax and fulfillment engine behind Shopify Markets Pro (now Shopify Managed Markets). Scala, Akka, Play, PostgreSQL, AWS, Kubernetes.
• Took it from proof of concept to $50M+ GMV across 3,500+ merchants and 136 international markets.
• After the Global-e acquisition, merged Flow’s stack into the joint product: catalog, currency, tax, order routing.
• Tuned for Shopify Plus Black Friday / Cyber Monday traffic.

Netsmart · Software Architect Team Lead

Software Architect Team Lead, Nov 2020 – Jul 2021 · Lead Software Architect, Jun 2020 – Nov 2020

• Led the team modernizing the flagship Avatar platform (a 40-year-old application) onto Java 11, Spring Boot and Angular.
• Mixed persistence across PostgreSQL and InterSystems Caché/MUMPS, with language translation tooling to move legacy code.
• CI/CD and containers: Docker, Kubernetes/OpenShift, Jenkins, JMeter.

Perfumania · Senior Applications Engineer

Jan 2018 – Feb 2020

• Owned perfumania.com end to end, from content publishing to order fulfillment.
• Migrated the legacy IBM WebSphere Commerce stack to Shopify (dotcom and Amazon Marketplace).
• Linux, Apache, deployments; integrations in Java, Mule, SQL Server and DB2.

Bug Labs · Lead Software Engineer

Feb 2008 – Apr 2012

Grew from QA to development lead of a 7–9 person team.

• SDK and rootfs for the BUG modular embedded Linux platform: OMAP kernels, Linaro/OpenEmbedded upstream work, drivers.
• Java/JNI/C APIs for hardware modules; XMPP and chunked-HTTP cloud messaging.
• DARPA street-level capture pilot (OpenCV, Python); Processing color-tracking demo at MoMA’s Talk to Me (2011); IoT data service backend (now dweet.io).

Earlier

Senior Software Engineer, Open-Xchange (2016–2017) · Senior Software Engineer, CooCoo (2015–2016; backend and middleware: Java 8, Apache Camel, AWS SQS/CloudWatch, MySQL, WildFly) · Software Engineer, Platform & API Services, TheLadders (2012–2015; Storm, RabbitMQ, Scala, Java, Clojure, Couchbase, Elasticsearch; metrics infra, CI, email migration ETL, data access for data science).

projects

local-llm-pi5 · 2026. Raspberry Pi 5 + Hailo-10H vs. Apple Silicon with measured tok/s. Hailo wins on compiled vision/Whisper models, not agentic LLMs (2B ceiling, 2k context).

publications

• Grim, Selinger, Braynen, Rosenberger, Au, Louie, Connolly. “Reducing Prejudice: A Spatialized Game-Theoretic Model for the Contact Hypothesis.” Artificial Life IX, MIT Press, 2004, pp. 244–250.
• Grim, Selinger, Braynen, Rosenberger, Au, Louie, Connolly. “Modeling Prejudice Reduction: Spatialized Game Theory and the Contact Hypothesis.” Public Affairs Quarterly 19(2), 2005, pp. 95–125.

education

Stony Brook University · B.S. Computer Science.

skills

Leadership & architecture: Technical Leadership, Solutions Architecture, API Design, Data Modeling, Stakeholder Management, Program Management.

AI / LLM: Generative AI, Retrieval-Augmented Generation (RAG), LLM Evaluation, AI Agents, Model Context Protocol (MCP), Claude API, Azure OpenAI, Azure AI Document Intelligence, Amazon Bedrock, pgvector.

Engineering: Python, TypeScript, Scala, Terraform, Kubernetes, PostgreSQL, Playwright, CI/CD, Embedded Linux, OpenCV.