Selected Work

AI systems that turn messy ops into pipeline.

Two GTM/RevOps engineering builds from a contract with a national 3PL. Built end-to-end with Claude as the execution layer.

Chris Cousins · GTM / RevOps Engineer TypeScript · Vercel · Neon · Claude · HubSpot
Company Intelligence Engine

Scoring every account on Volume, Fit & Timing — automatically.

Multi-source enrichment + scoring pipeline · serverless

The problem

Thousands of companies sat in the CRM with no signal on whether they were worth pursuing. Qualifying them was a manual, 95-point research SOP — accurate but unscalable, and impossible to run across the whole database.

The approach

  • Pull firmographics + traffic + ecommerce signals from multiple sources
  • Let Claude infer fit where data is thin or conflicting
  • Score each account on three gates: Volume, Fit, Timing
  • Tag into industry "pods" for heat-mapping & campaign planning
SimilarWeb
Traffic / volume screen
→
Apollo
Firmographics + contacts
→
Store Leads
Ecommerce tech + brand data
→
Claude
Inference + V/F/T scoring
→
HubSpot
Source of truth + pods
3
Volume / Fit / Timing gates scored on every account
Manual → Auto
A 95-signal research SOP turned into a repeatable engine
Pod heat maps
Industry-pod tagging drives campaign targeting
TypeScriptVercel ServerlessClaude / Anthropic SDK ApolloSimilarWebStore LeadsHubSpot APIGitHub Actions
Onboarding Genie

Turning sales calls into structured CRM data — on its own.

AI transcript-to-CRM agent · serverless + Slack

The problem

Every onboarding call held the details that actually matter — SKUs, volumes, timelines, requirements — but that context lived in recordings and never made it cleanly into the CRM. Manual note-entry was slow, inconsistent, and lossy.

The approach

  • Auto-scan the Drive folder where call recordings land
  • Parse transcripts, then have Claude extract structured fields
  • Write the results into HubSpot as 26 custom properties + a Neon record
  • Expose a natural-language "ask" endpoint so the team can query it
Drive recordings
Apps Script scan
→
Transcript parse
mammoth / text
→
Claude
Structured extraction
→
HubSpot + Neon
26 props · queryable
26
Structured HubSpot properties populated from each call
0 manual entry
Transcript → CRM with no human re-typing
Ask it anything
Natural-language query over the enriched data
TypeScriptVercel ServerlessClaude / Anthropic SDK Neon PostgresGoogle Apps ScriptHubSpot APImammoth