Case study 03
AI workflows · SaaS
AgentSDR
A cold-email tool for B2B teams. It finds and enriches leads through Apollo, reads your website for context, writes the outreach, sets up mailboxes, handles Gmail replies and bounces, syncs with Smartlead and keeps CRM status up to date.
My role
I architected most of it and built the bigger pieces with the team: onboarding, knowledge ingestion, the AI outreach, email delivery, mailbox operations and Langfuse tracing.
Mostly architected and built by me, together with the engineering team, across a few repo migrations.
Core technologies
System path
A simplified view of the product workflow.
- 01
Lead discovery
- 02
ICP qualification
- 03
Knowledge ingestion
- 04
AI sequencing
- 05
Mailbox events
- 06
CRM state
01
Product overview
How a campaign works
- 01Find and enrich leads and companies with Apollo.
- 02Define your ideal customer and pull in your website content.
- 03Generate pitches, sequences and CTAs, then edit them.
- 04Classify replies by intent and plan the follow-up.
- 05Timelines, notes, snoozing and do-not-contact lists.
02
My work
Mailboxes and delivery
Most of my time went into the email side: provisioning domains, managing mailboxes, warm-up and sending queues, Gmail watch and history processing, replies, bounces, inbox organisation and Smartlead sync.
03
Observability
Tracing the AI
Every generation and classification is traced in Langfuse. When an email came out wrong we could look at the exact prompt, inputs and output, which made fixing prompts a normal part of the work.