Pino Build

Portfolio

Selected work — not the full list.

These six cases are a public sample. The numbers below cover the rest of the practice: live builds, years shipping, and agency deliveries that never appear on this page.

Most agency work ships under the agency’s name, so I rarely publish a live client URL. Use the stats for scale; use the cases to see how I scope, ship, and measure.

The practice in numbers

40+

production builds shipped

20+

agency deliveries not listed here

5+

years shipping live software

6

cases written on this page

Six cases I can walk through in public

Not a catalogue. Each card is one project I can describe without breaking an NDA.

Web application

60%

less time on reporting

Concept projectWeb application

SaaS Analytics Dashboard

Monday reporting, without the spreadsheet ritual

The situation

A 12-person B2B team spent Monday mornings exporting from five tools into spreadsheets. By Wednesday, nobody fully trusted the numbers.

What shipped

One live dashboard: CRM, billing, product, support, and ads in the same view, with the filters the team already used.

60%

less time on reporting

5→1

sources in one view

12

people on the team

The resultMonday reporting stopped being a half-day job. The page is now the source of truth.

Stack

ReactTypeScriptRechartsNode.jsPostgreSQL

AI integration

80%

queries answered automatically

DOC · 12

Personal projectAI integration

AI Knowledge Assistant

Answers from your docs, not from memory

The situation

Support kept answering the same product questions from scattered documents. New hires took weeks to sound sure of themselves.

What shipped

An internal Q&A over company docs, with answers that point back to the source page.

80%

queries answered automatically

Docs

as the only knowledge source

4 wks

from brief to internal launch

The resultMost repetitive tickets never reach a person. The team still owns the docs; the assistant just reads them.

Stack

Next.jsOpenAI APISupabaseTailwindVercel

Custom system

14

manual flows automated

Concept projectCustom system

Business Automation System

Fourteen handoffs, one ops screen

The situation

A logistics SME ran 14 handoffs on email and spreadsheets. Mistakes showed up as wrong loads, not as tickets.

What shipped

Those 14 flows, automated: a queue, retries, and a simple screen for the exceptions that still need a person.

14

manual flows automated

20h

saved each week

<1%

error rate, down from 8%

The resultThe team only touches the cases that actually need judgment.

Stack

PythonFastAPIPostgreSQLReactCelery

Website

10 days

Figma to production

Agency projectWebsite

SaaS Marketing Website

Their brand. Their client. My hands on the build.

The situation

An agency had a finished Figma file and a launch date. Their own team was already booked on retainers.

What shipped

A pixel-accurate marketing site, fully responsive, with trial signup wired to the product.

10 days

Figma to production

4.2%

free-trial conversion, month one

0

client-facing mentions of me

The resultThe agency delivered under their name. I never joined a client call.

Stack

Next.jsFramer MotionContentfulTailwind

Custom system

70%

less back-and-forth

Personal projectCustom system

Customer Portal

Status without another email thread

The situation

Client updates lived in long email threads. “Where are we?” ate the afternoon.

What shipped

A self-service portal: invoices, project status, file drop, and a short thread per job.

70%

less back-and-forth

Self-serve

invoices and status

8 wks

including billing

The resultClients check the portal first. Email is for decisions, not for chasing status.

Stack

ReactSupabaseStripeTypeScript

AI integration

94%

extraction accuracy

Concept projectAI integration

Document Intelligence Tool

First pass by the model, last pass by a person

The situation

Lawyers spent hours pulling fields out of unstructured PDFs for each case.

What shipped

Upload, extract, review. Structured fields with a confidence score and a confirm step.

94%

extraction accuracy

3h+

saved per case

Human

still in the loop

The resultThe model does the first pass. A person only corrects the edge cases.

Stack

PythonAnthropic ClaudeFastAPIReact

How to read this page

01

This page is a sample

Most agency work ships under their name, so you will not see a live client URL. The six cases are what I can describe; the numbers above cover the rest.

02

Numbers over mood boards

I care what changed after launch — time saved, conversion, error rate — not just that a screen looks finished.

03

Same shape as your project

Every case follows the way I’d work with you: a clear situation, a defined build, checkpoints, then a result you can check.

Building something similar?

Tell me the situation and the deadline. I’ll say whether I can help — and how I’d break it into stages.

Start a project →