Skip to content
Waterfall
Taking new engagements — Q3

Engineering that compounds

We embed senior engineers inside product teams at companies in the US and Europe. Fewer people, further along — and the systems hold up after we leave.

Companies shipped for
8
Countries
3
Median engineer experience
11yrs
Product of our own
1

Clients: Hoag, Learned Hand, Perch Insights, SanarFlix, Veteran Benefits Guide, Alokai, Flowerplot, Berimbau.

01Clients

Companies that already had engineers

They didn't need more headcount. They needed the specific problem solved, by people who had solved it before.

  • Health systemUnited States · 2023
  • Legal technologyUnited States · 2024
  • Revenue operationsUnited States · 2024
  • Medical educationBrazil · 2022
  • Veterans servicesUnited States · 2023
  • People intelligenceEurope · 2025
  • Marketing analyticsUnited States · 2025
  • HospitalityUnited States · 2024

Several engagements are covered by NDA and aren't listed here. We can talk through the shape of that work — the problem, the stack, what changed — without naming the company.

02Studio

A small team, deliberately

Waterfall is a technology consultancy, and a deliberately small one. We take a handful of engagements at a time and staff each with engineers who have already built the thing you are about to build.

We work out of Rio de Janeiro, Brazil, close enough to the American day and early enough for the European one. Everyone here has built systems and then stayed to maintain them, which is the part of the job that changes how you build the next one.

Based
Rio de Janeiro, Brazil
Coverage
UTC−3 · overlapping US & EU hours
Founded
2021
  1. 01

    Seniority over headcount

    Adding engineers to a late project is the oldest mistake in software and it still gets made every quarter. We would rather send three people who have solved your problem before than eight who will learn it on your budget.

  2. 02

    Write it down

    A decision that lives in one person's head is a decision your team relitigates in six months. Everything we conclude lands in a document your engineers can read, argue with, and reverse without us in the room.

  3. 03

    Replaceable on purpose

    The best outcome of an engagement is that you stop needing it. We plan the handover in the first week, which has cost us the occasional renewal and has never once cost us a reference.

03Capabilities

What we're actually good at

Five things, done well. Anything outside this list we'll tell you honestly that we're not your best option.

Senior engineers embedded in your team, shipping from week one.

We don't hand you a discovery deck. We join your standup, take tickets off your board, and open pull requests against your repo. Most engagements are two to four engineers running as a self-sufficient pod inside your product org.

  • 01/04Full-stack delivery
  • 02/04Your repo, your process
  • 03/042–4 engineer pods
  • 04/04Async-first, timezone-overlapped

Make the deploy boring and the bill smaller.

Build pipelines that finish, environments that reproduce, and infrastructure your team can reason about at 3am. We inherit systems as often as we design them, and we leave documentation that outlives the engagement.

  • 01/04CI/CD that finishes under 10 min
  • 02/04IaC & environment parity
  • 03/04Observability you actually read
  • 04/04Cost teardown

LLM features that survive contact with real users.

Retrieval, evaluation, guardrails, and cost control — the unglamorous parts that decide whether an AI feature ships or quietly gets rolled back. We instrument before we optimise and we benchmark against your data, not a demo set.

  • 01/04Retrieval & context design
  • 02/04Eval harnesses and regression suites
  • 03/04Latency and spend budgets
  • 04/04Human-in-the-loop workflows

Inherit the codebase nobody wants to touch.

A frozen release train, a framework three majors behind, the one service only one person understands. We map it, stabilise it, and migrate it in slices that keep shipping — no eighteen-month rewrite with a big-bang cutover at the end.

  • 01/04Incremental strangler migrations
  • 02/04Test coverage from zero
  • 03/04Dependency and framework upgrades
  • 04/04Knowledge transfer, written down

Technical direction without a full-time hire.

Architecture review, hiring calibration, roadmap sequencing, and the hard call on build-versus-buy. For teams that have engineers but not yet the person who decides how they work.

  • 01/04Architecture review
  • 02/04Hiring loops & calibration
  • 03/04Roadmap sequencing
  • 04/04Vendor and build/buy calls
04Selected work

Systems we shipped

Some clients we can name, some we can't. The problems are real either way.

012024

Learned Hand

An AI case-preparation platform judges will actually trust

Fact extraction, issue mapping and motion analysis over court filings — where every generated claim has to trace back to a verifiable citation in the source document, because the reader is a judge. Now on the bench in over 300 courtrooms, and the company was acquired on the strength of it.

judges using it
300+
outcome for the company
Acquired
  • TypeScript
  • Python
  • Postgres
  • Claude
View case studyAn AI case-preparation platform judges will actually trust
022024

Perch Insights

Real-time analytics over the whole customer journey

CRM, dialer, marketing and billing data unified into one model, with agents watching for anomalies and answering questions in natural language instead of waiting for someone to build a dashboard. It raised a three-million-dollar round after launch.

raised after launch
US$3M
  • TypeScript
  • Python
  • Snowflake
  • React
View case studyReal-time analytics over the whole customer journey
032022

SanarFlix

Brazil's largest medical-education platform, under exam-season load

Two hundred thousand practice questions, thousands of video lectures and AI tutoring, serving Brazilian medical students — with traffic that spikes hard around residency and ENAMED exam dates.

medical-ed platform in Brazil
#1
practice questions
200k+
  • TypeScript
  • React
  • Node.js
  • Postgres
View case studyBrazil's largest medical-education platform, under exam-season load
042023

Hoag

Digital patient experience for a regional health system

Care navigation across a dozen specialty institutes, virtual visits and imaging, for a health system that sees more than 450,000 patients a year — the kind of surface where an ambiguous button costs someone a real appointment.

05Products

We build for ourselves too

Consultancies that never ship their own product lose the instinct for it. So we ship.

01Our first product

Zenda

Private beta2025

Every clinic's WhatsApp in one inbox

Zenda is built for the people who run medical schedules. Connect the WhatsApp Business number of every clinic you look after, work all of those conversations from one screen, and keep the calendar and the appointment confirmations right beside the thread.

Open the Zenda page
zenda.waterfalltech.xyz
02

Next product in progress

Same origin as Zenda: something we needed on client work and could not buy. This one is about what a team inherits when the engineers who built the system move on.

06How we work

Four stages, downhill

No discovery phase that bills for three months and produces a slide deck.

  1. Week 1

    Read the system

    We read the code, the incidents, and the backlog before we write a line. You get a written map of what exists and where it hurts — useful even if you never hire us.

  2. Week 2

    Ship something small

    A real change in production inside the first two weeks. It proves the pipeline, the access, and the working relationship all at once.

  3. Month 1 →

    Run as a pod

    Two to four engineers operating inside your process — your board, your reviews, your on-call if you want it. Weekly written updates, no status theatre.

  4. Exit

    Hand back stronger

    Documentation, runbooks, and a team that can maintain what we built. The engagement ends; the leverage stays. We're happy to be replaceable.

Contact07

Tell us what's stuck

One email, a real reply from an engineer within two working days. If we're not the right fit we'll say so and point you somewhere better.

Taking new engagements — Q3UTC−3 · overlapping US & EU hours