Iuliu Pop at his coding workstation

Iuliu Pop · Waterloo, ON

I build AI-powered product features on production-grade foundations.

Full-stack engineer at Pixee, shipping agentic security-remediation software for enterprise teams.

Email meDownload resumeSee the work

Applied AI, shipped

AI-powered product features, built inside real product constraints.

An agentic platform in production

Resolution Center is an agentic AppSec remediation product. I led the rebuild of its core workspace, the company's northstar priority, shipping the scan-centric findings, triage, and fix experience enterprises like NTT and Tier-1 banks use daily.

Tool-using agents

At Sparkcraft, an AI agency I co-founded with two UWaterloo grads, I designed and built an agentic assistant for the City of Waterloo: a tool-using system over municipal information that could take actions and chain multi-step workflows, built for a competitive municipal bid.

Evaluation and reliability

I've built tool-using agents, LLM evals using existing evaluation harnesses, and a maintainer-memory code-review agent. The recurring lesson: the model is the easy part. Evals, degradation paths, and knowing when not to use an LLM are the work.

The fundamentals underneath

Production systems have to hold up.

AI-assisted or not, production systems have to hold up. At Pixee I've:

  • Built the user-platform REST API with GitHub OAuth/OIDC on Java/Quarkus, enabling single-tenant on-prem deployment for NTT.
  • Eliminated N+1 query patterns in the findings data layer, collapsing ~11 API calls per page load into single scan-based endpoints.
  • Co-led PR Refresh across four SCMs (GitHub, GitLab, Bitbucket, Azure DevOps), decomposing a 13-issue MVP so re-scans update stale fix PRs instead of flooding repos.
  • Shipped a P0 CSV/JSON export honoring live filter state to unblock a Tier-1 banking customer.
Hypha observability architecture diagram
Hypha logo

Observability roots

Hypha

Before that, I co-created Hypha, an open-source observability tool for debugging distributed systems by correlating logs and traces: OpenTelemetry, Jaeger, Loki, Elasticsearch, Grafana, Docker, AWS ECS. You learn what production reliability means by building the tools people reach for when it breaks.

Read the Hypha case studyBuilding something like this? Email me.

The longer path

Curiosity, craft, and patience.

Before software, my path wound through meditation retreats, almost becoming a monk, cabinet-making, and freelance websites, then Launch School's deliberately slow route through programming fundamentals. I keep that unlikely path in the work: curiosity, attention to craft, and the patience to do things properly.

Where I fit

Applied AI engineering on real operational problems: agent workflows, natural-language interfaces, evaluation, and the product engineering that turns prototypes into software customers rely on.

Stack

TypeScript, React, Java, LLM workflows and agents, OpenTelemetry, Docker, AWS.

Contact

Based in Waterloo, ON.

iuliu.laurentiu.pop@protonmail.comResume PDFgithub.com/iulspoplinkedin.com/in/iulspop