Available for new systems

I build the private tools teams wish existed

I build browser-native tools that take the repetitive work off your team and keep the final call with the person who owns it. Messy process in, a calm private product out.

See how it works Get in touch
Private internal tools Browser-native Human-in-the-loop
private-ops · dashboard
⬡ Overview
◈ Ingest Queue
◎ AI Review
◇ Validation
○ Actions
306 processed today
98.4% accuracy rate
3 min avg cycle time
context-ingest✓
metadata-map✓
ai-handoff●
human-review·
validation queue
sync actions
◈ workflow-architect.js running
Hours become minutes The loops your team runs by hand, rebuilt as pipelines that run themselves
Human-in-the-loop Every system keeps the final call with the person who owns the decision
No training deck Interfaces people understand on the first open, because the tool fits how they already work
Modular, not fragile Built as small, inspectable modules, not a one-off script nobody can maintain later
How I work

I take the repetitive loop off people, and leave them the decision

Most repetitive work hides the same drag in small steps. You find the right context, move it into another tool, check the result, then push a decision back into the workflow. None of it is hard. All of it is time.

I rebuild that loop so it runs on rails and stays inspectable. Context is routed for you, the model stays visible while it drafts, and the check at the end is a control you own rather than one more thing to do by hand.

01 Gather the context The right inputs arrive on their own, so nobody chases screens for them.
02 Hand it to the model Messy context goes in under tight rules, structured help comes back.
03 Keep the decision human The final call stays visible, and it stays yours.
Capabilities

What I actually build

Where messy operations turn into something you can run and trust.

AI workflow orchestration

Structured handoffs where the model does the drafting and the person still makes the call.

Human-in-the-loop

Low-code automation

Automations that move work between your forms, files, and tools, so no one babysits the handoff.

MakeZapier

Browser workflow systems

Small browser extensions that live in the tab where the work already happens.

Chrome extensions

Metadata aggregation

Scattered fields and files pulled into one structured, queryable shape.

Pipelines

Review and QA tooling

Less repetitive review for the team, with the judgment and the audit trail left intact.

QA systems

Internal dashboards

Clean private interfaces built for the people who actually operate the system.

UI design

Human-in-the-loop AI

Every AI output stays inspectable and reversible, so a wrong call is easy to catch and undo.

Control

Process design

Messy real-world workflows mapped into a system people actually want to open each day.

Product thinking

Some of the best tools cannot be public

Sensitive work needs careful systems. Access stays scoped, every action leaves a trail, and a person signs off before anything commits. The point is safer execution that does not slow the team down.

Architecture

An AI-assisted review workflow

A private system that pulls the context together, hands it to the model under fixed rules, checks the structured result, and keeps the final decision with a person. I build these end to end, so the AI becomes something the team can rely on rather than a shortcut they have to second-guess.

⬡
Raw Input
context ingest
→
◈
Mapping
metadata map
→
◎
AI Handoff
structured prompt
→
◇
Human Review
final decision
→
○
Sync Action
controlled output
[TASK] Review Queue Snapshot
Manual Open. Copy. Paste. Retry.
PSD
Source File font metadata · v3_FINAL
—
XLS
Review Log metadata export · 247 rows
—
TTY
Field Entry
font_weight:
—
✕
Decision awaiting correction
blocked
Automated Ingest. Parse. Verify. Ship.
⬡
Context Ingest source + log + brief
—
◈
Metadata Map required fields · structured
—
◎
AI Verdict structured evidence grid
—
○
Sync & Approval correction patched · queued
—
✓ approved
Process

From messy workflow
to private product

01

Map the human workflow

Before any code, I watch what people actually do. Where the time goes, where the errors creep in, and which calls need real judgment rather than pattern matching. That map decides everything after it.

02

Find what to automate

The loops, the copy-pastes, the manual checks. I separate what a model can speed up from what still needs human eyes, then draw the line between them on purpose.

03

Build the AI handoff

Prompts return machine-readable output. A validation layer catches what looks off. The human review step stays fast and visible, and it is never optional.

04

Ship it as a real product

Not a script with a README, a real product. Clear state, honest error messages, an interface the team opens every day without being told to.

Reliability

Built in modules
Not fragile scripts

Every tool starts as a map of the human workflow, then breaks into small parts you can open one at a time. Context, validation, the AI handoff, the human review, the final action. When one needs to change, you change that one.

⬡ context-ingest.js 2.4 kb
◈ metadata-map.js 3.1 kb
◎ ai-handoff.js 1.8 kb
◇ validation-loop.js 4.2 kb
● human-review.js 2.9 kb
○ sync-actions.js 1.6 kb
core.js is ready to load modules
Toolkit

Where the work
actually happens

The tools I reach for, and what each one actually does for you.
Not into the jargon? Flip any card for the plain-terms version.

Experience

The work behind the work

Automation Systems Designer Ongoing

I build private tools for the repetitive work operations teams get stuck with. A model does the reasoning, a low-code layer moves the data, and the whole thing ships as a product a person actually wants to use. Most of it lives under NDA.

AI handoffsBrowser toolingInternal products
Metadata & Review Tooling Specialist Private clients

Review interfaces and metadata pipelines for high-volume QA, built to cut the manual labor without cutting the oversight. The validation logic and the AI review steps keep a human reviewer fast and firmly in control.

MetadataQA systemsHuman review
Internal Tools Builder Freelance

The small, specific tools operations teams never get from a product roadmap. Chrome extensions, automation dashboards, validators, context panels, whatever closes the gap that week.

Chrome extensionsDashboardsContext systems
Skip the sales deck

Let's make things efficient

If your team is losing hours to repetitive review or metadata work, that is the kind of problem I close. A calm interface over messy work, with the AI kept where it earns its place and the final judgment left with you.

Email me Connect on LinkedIn
amazing@pietro.works