Marketing software

// The problem

An agency builds every client report by exporting from ad platforms, analytics and CRMs into spreadsheets, and clients wait until the end of the week to see how a campaign is going.

// What we would build

A campaign dashboard that pulls spend, reach, clicks and leads from every platform the agency uses into one model, and a branded client portal where each client sees their own campaigns, notes and reports at any time.

// What is in it

// Stack

• Next.js
• PostgreSQL
• BigQuery
• Ad platform and analytics APIs
• Google Cloud

// The problem

A brand team uses general AI tools for product copy, emails and social posts, and every piece is rewritten because it sounds nothing like the brand.

// What we would build

A custom model fine-tuned on the brand's own approved copy, style guide and product data, so first versions read like the brand, with a review queue where brand managers approve, edit or reject and their edits teach the next version. It runs on GPUs the client controls, so unreleased campaigns stay private.

// What is in it

// Stack

• Fine-tuned open-weight model
• vLLM
• Python
• Next.js
• DigitalOcean GPU

// The problem

An advertiser, its agency and a group of publishers each report different impression and click counts, and invoices sit unpaid for weeks while the difference is argued.

// What we would build

A private permissioned ledger the advertiser, agency and publishers each run a node of, where daily delivery figures from each party are recorded and matched, with a portal that flags differences while they are small. If everyone already trusts one ad server's numbers, keep them; the ledger is for when nobody does.

// What is in it

// Stack

• Next.js
• PostgreSQL
• Hyperledger Besu (private, permissioned)
• Ad server reporting APIs
• Google Cloud

// The problem

A brand runs community campaigns in Telegram and Discord, tracks referrals and tasks in a spreadsheet, and pays out rewards to bots and duplicate accounts.

// What we would build

A bot for Telegram and Discord that issues each member a referral link, tracks tasks and referrals, checks for duplicate and automated accounts before any reward is granted, and gives the marketing team a dashboard of the campaign as it runs.

// What is in it

// Stack

• Node.js
• Telegram Bot API
• discord.js
• PostgreSQL
• Next.js
• DigitalOcean

// The problem

An agency exports spend and results from each ad platform, analytics tool and CRM into spreadsheets for every client report, and the numbers rarely match from one month to the next.

// What we would build

A data warehouse that loads ad platforms, analytics and CRM data for each client on a schedule, models them with one set of definitions, and feeds the dashboards and reports your team already sends, with each client seeing only their own data.

// What is in it

// Stack

• PostgreSQL
• dbt
• Airbyte
• Looker Studio
• DigitalOcean

// The problem

An agency runs campaigns for clients in Canada and Latin America, and every landing page and email is copied into documents, sent to translators by email, and pasted back by hand.

// What we would build

A localisation workflow that sends campaign pages and emails from your CMS to translators and native reviewers, keeps each client's glossary and brand voice, checks layouts in every language before launch, and lets a client approve each language in one place.

// What is in it

// Stack

• Next.js
• PostgreSQL
• Crowdin
• Claude API
• DigitalOcean