I take technical products to market, and build the systems that show what’s working.

Positioning

For each assistant, I turned the capability into a use case for a specific buyer: manufacturers, distributors and eCommerce teams. For Shopping, I positioned agentic commerce as a product discovery problem first, not a checkout change. Shoppers may start in an AI assistant instead of a category page, so product data, search and recommendations have to be ready for that. A readiness audit made that problem concrete and gave prospects an easy first step.

What I built for the Shopping launch

• The Agentic Commerce Audit offer, with a HubSpot form and promotion workflow.
• The Agentic Commerce Guide page on Webflow.
• A 6-email Salesloft sequence to 353 contacts, with a 150-per-day send cap and a holiday-aware schedule.
• A 3-message LinkedIn flow in HeyReach.
• A paid LinkedIn thought-leadership concept with retargeting to the audit.
• Suppression logic so prospects already invited to the webinar weren't pitched it again.

What this shows

Product marketing end to end: positioning by buyer, one offer, and four channels that stayed consistent with each other, down to send caps and suppression rules.

Constraints

No BI budget. No direct Salesforce connection.

What I built

• n8n workflows that pull data from each tool through its API on a schedule.
• An executive dashboard in SharePoint, built with Claude Code, covering paid, outbound, website and CRM performance.
• Opportunity stage read from the HubSpot property that the Salesforce sync writes to, so the dashboard never touches Salesforce directly.

Key decisions

• SharePoint instead of a separate BI tool, so leadership opens it where they already work.
• Switched Instantly to its campaign analytics endpoint, which includes bounce data and cut the workflow to one API call.

Proof

• 30+ team hours saved every week.
• Backfilled late July to early November 2026, and totals matched the source tools exactly.
• Checked every scorecard against raw data, which caught a duplicated test date and a filter tied to the wrong source before anyone made a decision on them.

What I built

• An n8n webhook that creates a HubSpot contact only when a LinkedIn connection request is accepted.
• Sets Lead Source to LinkedIn and Lead Source Detail to HeyReach on new contacts only, so existing attribution is never overwritten.
• Finds or creates the company by email domain, and skips personal email domains.
• Adds every accepted lead to a static list for follow-up.
• A backfill workflow with a dry-run toggle to replay historical connections safely.

Key decisions

• Trigger only on accepted connections, with an event check added after HeyReach's webhook started sending the wrong event type. Bad events can't create junk contacts.
• Write lead source on new contacts only, so first-touch attribution from other channels survives.
• Keep these contacts non-marketing and out of Salesforce until they engage. This holds down HubSpot marketing-contact costs and keeps unqualified records out of sales data.

Proof

• 200+ contacts backfilled from 4 months of outreach, with dry-run output checked against HeyReach before the live run.
• We count ICP-matched accepted connections as qualified leads.
• Across the LinkedIn campaigns feeding it: about 20% of connection requests accepted, 7% replied, and 10+ meetings booked.

Constraints

Read-only Google Tag Manager access. No changes allowed that could affect live visitors.

What I did

• Reproduced the freeze with local browser testing instead of changing the live site.
• Traced it to the Google Ads tag's automatic content scanning, which hung on an oversized ASP.NET ViewState field present on every page.
• Proved it: removing the field removed the freeze. Ruled out hosting and the Tag Manager configuration.
• Handed off cleanly: the site fix went to the developer, and the tag-setting fix was written up for the client.

Why it matters

Marketing tags can break the pages they're meant to measure. I can debug at that layer instead of escalating blind.