BizOps and GTM at Shippo
Four threads from my time at Shippo, a mid-stage shipping platform: partner program ops, carrier claims recovery, USPS reconciliation, and the AI tool governance pilot. Each one has a working artifact you can click into, so you can see what I actually built.
A roundup of the work I've shipped at Shippo on the BizOps and GTM side. Each project below has its own case study. I wrote them while the work was actually happening, so they're closer to working notes than a polished retelling.
Skim the four cards to get a feel for each project. Then open any case study to see the actual dashboards, Slack flows, and Wiki pages I built. The look there matches the actual artifacts, not this portfolio's styling.
The four case studies
Each card links to the actual case study. These are the things I wrote during the work, not a polished retelling after the fact.
The Partner Program backbone
I took over a partner program that had real revenue but nothing structured underneath it. So I built the cadence, the source of truth, and the reporting layer it needed: a leadership reporting redesign, a Confluence hub rebuild, a Databricks partner-health dashboard, and the first AI-powered workflows in BizOps.
GSR claims recovery dashboard
A working dashboard for carrier service-failure refunds (late deliveries, lost packages, damaged shipments across UPS, USPS, FedEx). Eligibility scanning, auto-filing through Zendesk macros, win-rate tracking, denial-reason analysis, recovery trend monitoring. The whole thing is set up so a claims operator can answer "what's my MTD recovery right now?" in two clicks.
USPS unmanifested & duplicate label reconciliation
Carrier surcharge disputes used to live entirely in someone's head, and we were leaking money because of it. I reconciled raw USPS data against four internal systems, codified the joint review process into a 5-phase SOP, and built the four artifacts that made the whole thing repeatable. Cycle time went from three weeks to three days.
GTM AI tool governance pilot
This kicked off after I noticed two teams had quietly started building overlapping AI tools (a partner recommender and a Databricks predictive model). I built a lightweight governance system to catch that earlier: one rule, three tiers, a 5-minute intake card. The pilot is live with 4 to 5 GTM users, and we surfaced 11 tools in the first week alone.
How these fit together
Four projects, but they all started in the same place. The bottleneck before each one wasn't capability. It was indecision and people quietly duplicating each other's work. So my job ended up being less about building the thing, and more about building the system that makes the next person's version of the thing happen faster.
Replace the work, don't just speed it up
Manual reporting cycles, ad-hoc SQL pulls, hand-keyed reconciliation. I tried to replace each of these with a system, not just a faster way to do the same thing by hand.
Lead with the decision someone actually needs to make
Leadership emails got the R/Y/G status and the asks pinned to the top. Dashboards answered "is this growing?" in 60 seconds. AI governance tier checks decided sign-off depth before anyone spent time scoping.
Codify the second time it happens
The first carrier dispute was a one-off. The second became a 5-phase SOP. The first AI-tool collision became a registry. The first partner deep-dive became a Lakeview dashboard with 27 datasets and a Python build script.
Ship the rough version, then make it nice
Every case study above started scrappy. The dedup tool was a Streamlit script. The leadership update was a Google Doc. The polish came after the value was already proven on real cycles.