---
title: "Atlas | Dylan Tullberg"
description: "Commercial intelligence, from company research and qualification to controlled CRM delivery."
url: "https://tull.ink/work/atlas"
---

[← All projects](https://tull.ink/work)

Aquablu · Commercial intelligence · 2026

# Atlas

Making commercial criteria hold from research to sales.

More prospect data does not automatically make better sales work. Atlas connects company research, buyer qualification, and outreach while keeping incomplete records from becoming someone else’s problem.

My responsibility

I built enrichment workflows, identity and qualification controls, researched outreach, CRM delivery, and operator interfaces as part of the wider Atlas project.

## Follow the criteria, not just the record.

The ideal customer profile (ICP) guides company fit, relevant offices, buyer roles, and research. Different evidence leads to different next steps.

**ICP alignment**Geography Company fit Office scope Buyer roles Evidence

-   Business data
-   Company websites
-   Professional profiles
-   Location sources
-   Contact enrichment

**Example Co.**Identity resolved

**Office A**

**Office B**

**Workplace lead**Role, location, contact

**Sustainability lead**Research + talking points

**Buyer role missing**Further discovery ↻

**Needs more research**Not ready for delivery

**CRM-ready contact**Generic outreach when eligible

**Personalized outreach**Supported by research

**Sales feedback**CRM status · Campaign activity · Replies

Inform the next cycle

**Review and refine**People interpret the feedback

Provider capacity · Retries · Recovery · Controlled delivery

Illustrative workflow with example data.

## Three decisions that shape the work.

### Resolve identity before adding more data.

Uncertain imports need investigation before they become trusted company records. I built identity workflows to keep those questions visible through research and reconciliation.

### Implementation detail

Matching, merging, provisional records, and identity reservations support reuse of existing companies and competing claims during concurrent processing. Geography, company criteria, offices, and buyer personas guide selection.

### Separate a reachable contact from a researched approach.

An email address alone is not enough for a researched approach. My implementation checks contact eligibility separately from readiness for personalized outreach.

### Implementation detail

Company context, buyer profiles, and substantive talking points support personalization. Eligible contacts can follow a separate generic-outreach path. Unresolved records stay available for further work instead of silently becoming sales work.

### Make a complex pipeline operable.

Operators need to configure workloads, inspect progress, and recover incomplete runs. I built the controls and interfaces for those tasks.

### Implementation detail

Adaptive processing accounts for database pressure and provider capacity. Retries, alerts, progress tracking, and controlled HubSpot synchronization keep failures inspectable. I also contributed to the second-generation processing architecture and separation from the legacy engine.

## A workflow sales can use and operators can inspect.

The implementation connects research, qualification, outreach preparation, and controlled CRM delivery. Operators can see what progressed, what needs more work, and where processing needs intervention.

CRM lifecycle and lead status, campaign activity, and replies return to Atlas records. People review that feedback to inform the next cycle; it is not automatic retuning of the ICP.

TypeScript · React / Next.js · PostgreSQL / Supabase · Inngest · HubSpot

[Explore more work](https://tull.ink/work)[Discuss a role or engagement](https://tull.ink/#contact)