Specific problems, measured results.

Context

A stealth pre-seed company ran its outreach through an AI pipeline: find prospects, find and verify contacts, draft emails, publish content.

Problem

The prospecting stage rarely found a contact and most runs failed, so very few leads reached outreach and each one cost too much.

What we built

• A rebuilt prospecting agent with contact verification
• Outreach drafted under style rules enforced in code
• Every model call costed; a kill switch that pauses the pipeline

Context

A direct-to-consumer brand ran its operations across four systems: a storefront, a spreadsheet, a supplier inbox and a bank app.

Problem

Nobody could say what one order made after costs or how much of the supplier prepayment was left, and expenses were logged by hand.

What we built

• Net profit and margin per order, after cost of goods
• A running supplier prepayment balance
• An expense log kept in sync with the team’s spreadsheet
• AI drafts of blog and social posts, published only after approval

Context

The treasury team of a Series B fintech, funding payouts through many providers and currencies every day.

Problem

Balance, liquidity and reconciliation checks ran by hand from spreadsheets: slow, and easy to get wrong.

What we built

• A balance dashboard that flags providers at risk of running out of funds
• Automated reconciliation of receivables and refunded transactions
• Treasury logic moved from spreadsheets into tested dbt models

Context

The payments and customer operations of a European neobank, a small team manually processing data for millions of customers.

Problem

Customer records were cleaned and filtered by hand, and the team needed reporting it could explore without asking an analyst each time.

What we built

• SQL automation in Snowflake for cleaning and filtering customer records
• Process automations, built with the data engineers and product managers
• Looker dashboards on Snowflake data