Results, not promises.

Court deadlines that don't depend on anyone remembering.

Law firm · Size: Mid-size firm with an active caseload

The bottleneck we found

Court deadlines tracked by manually checking ADVBOX (a Brazilian legal practice management platform). Someone had to open it, read every new docket update, judge whether it triggered a deadline and alert whoever was responsible. Daily, repetitive work with one unacceptable failure mode: a single day without checking could mean a missed deadline — the most expensive mistake a firm can make.

What we built

A platform that syncs deadlines, court notices and docket updates straight through the ADVBOX API. Each update is read and summarized by AI, which classifies its urgency. The person responsible is notified by WhatsApp and email, prioritized by how close the deadline is. A monorepo with a Next.js dashboard, a Fastify API and asynchronous BullMQ workers; PostgreSQL on Supabase, Redis, magic-link sign-in, validation of every external input, and Sentry monitoring.

A monthly close that stopped eating days.

Nonprofit organization · Size: Multiple bank accounts and a recurring monthly volume of entries

The bottleneck we found

Month-end close done by hand: download each bank statement, classify entry by entry, check receipts, assemble the report. A long process, dependent on one person, carrying the classic risk of manual reconciliation — duplicate charges and misclassified entries that only surface months later.

What we built

A system that automates the whole path, from statement to approved financial report. It imports statements from several banks with rule-based classification and a dry-run cycle before anything is committed. It issues consolidated payment slips and settles them automatically through the gateway webhook, with every operation idempotent by source key — the same charge never lands twice, even if the webhook repeats. AI-assisted receipt capture proposes account, category, date and amount side by side with the file, for human review. Reports with custom month ordering and spreadsheet export. React + TypeScript on Supabase, with Row Level Security, edge functions and a mirrored test environment holding no sensitive data.

A dental practice with charts, scheduling and finance in one place.

Dental practice · Size: Practice with multiple chairs and a team of clinicians

The bottleneck we found

The practice's information scattered everywhere: charts on paper, images on somebody's computer, the schedule in a spreadsheet, finances in another. Nothing talked to anything. Finding a patient's history meant looking in three places, and the owner had no consolidated revenue figures by clinician or by procedure.

What we built

A complete practice system. Patient charts with an interactive dental chart, clinical history, configurable anamnesis questionnaires and automatic alerts based on the answers. Document upload in several formats, including DICOM, with viewing and download. Scheduling by clinician and by chair, with status tracking. Inventory management with low-stock alerts and supplier records. A finance module with monthly income statement and trial balance exported to PDF. React + TypeScript with PWA support, a Node.js back end and PostgreSQL.

A sales operation that moved off the spreadsheet.

B2B sales operation · Size: Sales team running active prospecting

The bottleneck we found

Prospecting tracked in a shared spreadsheet. Nobody knew precisely which stage each opportunity was at, or who owned it. The list was exported and re-imported by hand, with diverging versions circulating.

What we built

A sales hub with a drag-and-drop kanban prospecting pipeline, an SDR dashboard, scheduling, request management and an admin area with role-based access control. Validated forms and spreadsheet export for whoever still needs it. React + TypeScript on Supabase.