I'm an engineering leader who still codes. I build platforms, operating models, and teams — the kind that keep delivering long after the initial investment.
Currently leading the Carrier Integrations Division at Shippo, where I designed and built CIF (Carrier Integration Factory) — a 7-phase AI-driven delivery methodology where LLMs handle 80–90% of spec generation, engineering scaffolding, validation, and rollout. Before that, I spent 4.5 years at Loadsmart building Kraken from scratch: an any-to-any integration platform that grew from a cost-reduction initiative into foundational infrastructure serving 139+ shippers and 12 carriers.
My thing is finding the system inside the problem — then building it so others can build on top of it.
What I'm working on
- 🏭 Scaling CIF (Carrier Integration Factory) — getting from Level 1 (Defined) to Level 2 (Standardized) with real throughput data
- 🤖 Exploring agentic orchestration with Claude Code — outer session as orchestrator, inner subprocess sessions per task
- 📐 Thinking about what "AI-native" engineering management actually looks like in practice
Stack I live in
AI tooling I actually use
A few things I've built
| Project | What it is |
|---|---|
| CIF | AI-driven carrier integration factory at Shippo. 7-phase lifecycle, LLMs in 5 of 7 phases. |
| cid-engineering-manager-assistant | AI tooling I built for my own EM workflows — perf reviews, sprint management, metrics. |
| shippo-safety-detection | LLM-aware security library. Built from scratch, integrated into the core backend. |
| Kraken | The integration platform I founded at Loadsmart. EDI → any-to-any → 139 shippers, 12 carriers. |
| GSD-Lite | Lightweight agentic orchestration framework on top of Claude Code. WIP. |
How I think about engineering
The most important question isn't "how do we solve this?" — it's "what system are we building that makes this class of problem easier next time?"
I've built two platforms that compounded: Kraken at Loadsmart (other teams built on top of it) and CIF at Shippo (other engineers use it to deliver faster). The pattern is the same: resist the surface metric, build the durable thing, prove the hypothesis yourself before scaling it.
| Metric | Count |
|---|---|
| PRs authored (last 27 months) | 130+ |
| PRs reviewed (last 27 months) | 186+ |
| Repos contributed to | 9 |
| Production incidents led (Sev1/Sev2) | 10+ |
| Follow-up completion rate | 100% |
| Engineers managed (peak) | 20+ |
| Carrier integrations delivered | multiple |
Find me
