Danilson Ramos · Zürich, Switzerland

Build with AI. Verify like an engineer.

Real builds, documented failures and verifiable fixes for moving from prototype to reliable software.

One accountable builder, from idea to production. Fixed scope, fixed fee — a SaaS v1 from CHF 20’000 to 45’000, in 4 to 6 weeks.

Discuss a projectSee production work
Currently booking

Reply within 24h, Mon–Fri.

AtamataxTax situation engine · document workflow · fixed-price fulfillmentHorveraDeterministic rules engine · claim workflow · programmatic search

BUILD OS · EVIDENCE GRAPH

Move from prototype to reliable product with visible checks.

Documented products can yield decisions, failure modes and reusable checks. The graph connects evidence to problems, patterns, builds and tools without publishing unmeasured statistics.

  1. 01Frame
  2. 02Build
  3. 03Verify
  4. 04Document
  5. 05Reuse
ShipCheck assessment · FR →Explore Build School →See the products being built · FR →The Build OS · FR →
  • Production SaaS
  • Deterministic rules engines
  • Stripe payments
  • Regulated workflows
  • Third-party APIs
  • Search-driven products
  • Atamatax · Horvera · Wabiro · SponsorsMetrics
  • 10 years shipping software
  • Zürich, Switzerland — remote across Europe

Selected systems

Five products I designed, built and operate, four of them taking real payments. Each one is a case study: the problem, the architecture, the hard decisions — and each one states its status.

Atamatax — US expat tax infrastructure
TaxFintech

Atamatax

US expat tax infrastructure

In production · payments live

Explore the systemVisit site
Problem
Americans abroad carry fragmented US obligations — FBAR, PFIC rules, Form 8621, catch-up procedures — hidden inside ordinary foreign accounts and funds; the market offers a specialist at several thousand dollars, or nothing.
System
Tax situation engine · document workflow · fixed-price fulfillment
Role
Research, product, rules engine, engineering, payments, search acquisition
Hardest part
Making uncertainty honest: every classification carries its source, a confidence level and the assumption still to confirm — and never reads as a legal determination.
Horvera — flight compensation infrastructure
ConsumerLegal rules

Horvera

Flight compensation infrastructure

In production · payments live

Explore the systemVisit site
Problem
A disrupted passenger does not know air law and should not have to; the market answer is a claims company that keeps 30 to 44% of the payout.
System
Deterministic rules engine · claim workflow · programmatic search
Role
Research, product, rules engine, engineering, payments, programmatic search
Hardest part
Three jurisdictions that share one text and diverge on amounts, currencies and the reading of delay — carried by one versioned engine that shows its trace and its sources.
Wabiro — identity and relationship tests built for search and sharing
ConsumerTests

Wabiro

Identity and relationship tests built for search and sharing

In production · payments live

Explore the systemVisit site
Problem
Personality and relationship tests are a huge search category built on forced sign-ups, endless quizzes and results with no method.
System
Test engine · versioned scoring · share loop · one-off payments
Role
Research, product, scoring systems, engineering, payments, SEO acquisition
Hardest part
Holding the privacy line — answers never leave the device — while still allowing sharing, head-to-heads and paid reports.
Vellaci — EU Cyber Resilience Act compliance for product manufacturers
B2BProduct security

Vellaci

EU Cyber Resilience Act compliance for product manufacturers

In production · subscriptions live

Explore the systemVisit site
Problem
The EU Cyber Resilience Act turns product security into questions with legal deadlines — what is exposed, what must be reported inside 24 and 72 hours, what evidence can be produced — and most manufacturers answer them from spreadsheets and ticket timestamps.
System
SBOM ingestion · vulnerability intelligence · deadline engine · evidence vault
Role
Research, product, regulatory modelling, engineering, payments, search acquisition
Hardest part
A reporting clock a regulator will accept: it starts when a named reviewer confirms awareness with a rationale — not when a ticket was created — and it is computed server-side with its rule version, source timestamp and timezone persisted, never only in the browser.
SponsorsMetrics — the pricing cockpit for YouTube talent agents
Creator EconomyPricing

SponsorsMetrics

The pricing cockpit for YouTube talent agents

Private beta · design partners onboarding

Explore the systemVisit site
Problem
YouTube talent agents in France price sponsorships by feel; when a brand says “too expensive”, there is no external comparable to answer with.
System
YouTube API pipeline · CPM benchmark · deal desk · subscription billing
Role
Product, engineering, design, pricing-data pipeline
Hardest part
Bootstrapping a data product: the benchmark is only worth something once real campaigns are connected — hence a private beta with design partners before open access.

All systems →

Also built

Evidence in production

Capabilities

Capabilities proven by systems.

Not a list of technologies: every row points at the production product where it is implemented.

  • Rules engines

    Deterministic, versioned decisions with a trace and sources — tax, air passenger law, scoring.

    Proven byAtamataxHorveraWabiroVellaci

  • Payments and billing

    Stripe: fixed-price packages, one-off purchases, subscriptions, signed webhooks, billing state in the database.

    Proven byAtamataxHorveraWabiroVellaciNeucelleBuild School

  • Regulated workflows

    Rules of law turned into journeys: obligations, thresholds, exceptions, what stays with a professional.

    Proven byAtamataxHorveraVellaciNeucelleComptaNexo

  • Case and operational workflows

    One file per case: documents, states, follow-ups, review before release, escalation.

    Proven byAtamataxHorveraVellaci

  • Complex data modelling

    Facts → obligations → forms; airports and licences → scope; campaigns → CPM distributions.

    Proven byAtamataxHorveraSponsorsMetricsVellaci

  • Third-party APIs and OAuth

    YouTube Data and Analytics in read-only OAuth, encrypted tokens, a normalisation pipeline per niche.

    Proven bySponsorsMetricsVellaci

  • Search-driven acquisition

    Programmatic pages generated from the same data as the engine, free tools, a result before any contact form.

    Proven byAtamataxHorveraWabiroVellaciComptaNexo

  • Authentication and permissions

    Server-enforced roles, revocable external-advisor access, rights granted file by file, enforceable MFA and SSO, account-scoped workspaces.

    Proven byVellaciAtamataxSponsorsMetricsNeucelle

  • Analytics and instrumentation

    Product events, a measured funnel, reports verified from an official API — not from a screenshot.

    Proven bySponsorsMetricsHorverabynilson.ch

  • AI-enabled workflows

    Extraction, classification, structured generation — after the deterministic engine, never in its place. None of the products in production depends on it today.

    No product proof — seeAI developerWhen a SaaS should use AI

How I work

Diagnose

20 minutes, free. Business model, workflows, risks and product scope. If I’m the wrong person, I say so.

You bring
Business goal, users and the current constraint.
You get
Fit decision and a written problem brief.
Typical time
20 minutes

Architect and propose

Within 48h: data model, system boundaries, integrations, UX — and one fixed fee with dated milestones. No “starting at”.

You bring
Requirements, examples, access to stakeholders.
You get
Scope, architecture, risks, milestones and fixed fee.
Typical time
Within 48 hours after diagnosis

Build

4 to 6 weeks of controlled iterations. A demo every Friday on a Vercel preview. You watch the system grow live — no black box.

You bring
Timely feedback and one decision maker.
You get
Working product, weekly demos and source code.
Typical time
Typically 4–6 weeks

Launch

Payments, analytics, deployment, operational workflows. You leave with the repo, the accounts, the pipeline — and 30 days of 24h replies.

You bring
Production credentials and launch approval.
You get
Deployed product, handover and 30 days of support.
Typical time
Included in the build plan

Improve

Measured iteration on real usage: a fresh fixed-fee scope per increment, or nothing if your team takes over. No hidden dependency.

You bring
Usage feedback and the next business priority.
You get
A measured next phase or a clean handoff.
Typical time
Scoped per iteration

Offers

Three product engagements. Nothing else.

I take few projects, chosen for what they change in a business. Every engagement is fixed-fee, on a written scope, with a demo every week.

01 — Product

Product Rebuild

For
A company whose digital product has aged, slows the team down or loses customers — or a prototype that has to become a production product.
Signals
  • — Every change breaks something else
  • — The business rules live in two people’s heads
  • — The prototype found customers, not the architecture to keep them
Deliverable
The product rebuilt around its domain model: data migrated, rules tested, payments, roles, admin, observability, deployed without downtime. You keep the repo and the accounts.
Format
From CHF 20’000 fixed fee · typically CHF 25’000–80’000 · 5–10 weeks.
Running in
Atamatax · Vellaci · Horvera
Not a visual refresh or a theme swap.
Check project fit

02 — Operations

AI / Automation System

For
A business that pays people to read, check, re-key or chase — documents, case files, rules, controls, double entry.
Signals
  • — Case files arrive by email and spreadsheet
  • — Every case follows rules someone checks by hand
  • — Deadlines and follow-ups are managed in a calendar
Deliverable
A workflow system: structured intake, a versioned rules engine, states and follow-ups, human review where it is due, an audit log. AI only where it beats a deterministic rule — extraction, classification — and always reviewable.
Format
From CHF 20’000 fixed fee · typically CHF 20’000–50’000 · 4–8 weeks.
Running in
Atamatax · Vellaci · Horvera
Not a chatbot plugged into your documents. None of the products in production depends on an LLM today: the proof here is the rules engines and the workflows.
Check project fit

03 — Acquisition

Growth Infrastructure

For
A product with an offer and customers but no measurable acquisition engine — search traffic, answer-engine citations, conversion, funnel.
Signals
  • — The site shows up neither in Google nor in AI answers
  • — Nobody knows which page produces customers
  • — The contact form receives noise, not projects
Deliverable
Technical SEO and programmatic pages from your data, AEO/GEO (entity graph, citable answers, llms.txt), free tools that qualify, an instrumented funnel, a scored intake and a conversion dashboard.
Format
From CHF 20’000 fixed fee · typically CHF 20’000–40’000 · 4–6 weeks.
Running in
Atamatax · Horvera · Wabiro
Not a marketing site, not content by the kilo, not paid ads.
Check project fit

The entry point: the diagnostic

CHF 2’400, 7 working days, a written deliverable — scope, architecture, risks, plan, budget. 100% credited against an engagement signed within 60 days.

Order the diagnostic

Minimum for a product engagement: CHF 20’000. Below that, I will tell you honestly what would serve you better.

What I do not take as a product engagement: marketing sites, WordPress or Shopify themes, hourly support, native mobile apps, staff augmentation. To redesign an existing site, there is Design Rescue (in French), at a fixed price; to fix bugs in an existing application, Code Rescue.

Services

Product engineering

I take few projects, and only three kinds: rebuilding a digital product that has aged, automating repetitive information or workflow work, building a product’s acquisition infrastructure.

All services →

Also

Notes & essays

SaaS architecture, rules engines, the economics of a v1, AI in production — written from shipped systems, not from the documentation.

All writing (14) →

FIELD NOTENotes de terrain · décisions et règles applicables →

About

Danilson Ramos — développeur web full-stack et designer produit à Zürich, Suisse
Danilson Ramos47.37°N · 8.54°E

FR (native) · EN (fluent) · PT (heritage) · DE · IT

Independent product engineer, in Zürich. Ten years writing software, five pairing it with design. I work mostly with founders and small companies who have a real operational problem — rules, workflows, money moving — and need it to become software.

I carry everything from architecture to production: data model, rules engine, interface, payments, email, deployment, observability. A SaaS, AI and full-stack developer — but the technology is the evidence, not the argument. The argument is the products I operate: Atamatax, Horvera, Wabiro, Vellaci, SponsorsMetrics.

I don’t subcontract. When the data model is shaky, I rewrite it in week two so you don’t pay for it in week ten.

Default stack

  • Next.js
  • TypeScript
  • Tailwind v4
  • Postgres
  • Vercel
  • Supabase
  • Stripe
  • Resend
  • API LLM
  • Python
  • PostHog
Who Danilson Ramos is

Currently booking — 1 to 2 projects per quarter

Prefer to build it yourself?

Explore Build School →

Frequently asked questions

  • What does Danilson Ramos do?

    Danilson Ramos is an independent product engineer in Zürich, Switzerland. He designs and builds production SaaS, custom software and AI-enabled products — in particular systems built around rules, workflows and payments. One person covers the architecture, the Postgres data model, the Next.js interface, Stripe, email, deployment and observability. Creator of Atamatax, Horvera, Wabiro and SponsorsMetrics.

  • How much does it cost to build a SaaS in Switzerland?

    Between CHF 20’000 and 45’000 fixed fee for a v1 built properly. That is also the minimum engagement: an internal tool, a portal or a workflow is built as an automation system, from CHF 20’000. The price is set after the scope call and does not move. The SaaS cost calculator places your project in five minutes.

  • How long does it take to build a SaaS v1?

    Four to six weeks of sprint after a week of scoping, with a demo every Friday on a Vercel preview. What a v1 must contain is detailed on its own page.

  • What is the difference between an MVP and a production v1?

    An MVP may still depend on you to work. A production v1 stands on its own: authentication, real payments, email, handled errors, an admin interface, monitoring. A prototype impresses in a demo; a v1 survives its first customer.

  • When should a SaaS use AI?

    When a task is done by hand on text or documents, when the error can be caught and when real examples exist to evaluate against — extraction, classification, structured generation. Not for a decision that must be justifiable, nor for what a query already does. See AI development.

  • What stack is suitable for a modern SaaS?

    Next.js (App Router), TypeScript, Postgres — Supabase when its managed parts save time —, Stripe, Resend, Vercel, and language-model APIs when AI is justified. It is the stack behind my own products and this site.

  • Freelancer or agency for SaaS development?

    A senior independent developer carries the architecture and the code himself, fixed fee, with one point of contact — cheaper and faster than an agency as long as the product fits in one person’s hands. Beyond that, I say so on the first call.

  • Are you available for a new project?

    I take one to two projects per quarter to stay focused. The fastest way in is the project-fit check — six questions, an immediate orientation, then a scope review. Otherwise, danilson.daveigaramos.pro@gmail.com — reply within 24h, Mon–Fri.

Have a product that needs to become a real system?

Six questions to find out whether your project is a production v1, an architecture audit or something else — then a scope review, free.

Check project fitEmail Danilson

danilson.daveigaramos.pro@gmail.com · Reply within 24h, Mon–Fri