The architecture behindevery decision The architecturebehind everydecision

We design and build the systems that connect a company's operations, its information and its decisions.

Approach

First we understand the operation. Then we build the technology.

Artificial intelligence, dashboards and automation don't work on top of disordered data. The foundation has to come first: clear processes, consistent data and systems you can rely on.

Nor do we start from the idea that everything has to be replaced. We assess what is worth keeping, what to integrate and what to build from scratch, and we turn fragmented operations into infrastructure that grows with the organization.

  1. 01

    Understand

    We survey the real operation — processes, systems, data and responsibilities — before proposing anything.

  2. 02

    Design

    We define the architecture: what stays, what gets integrated and what gets built from scratch.

  3. 03

    Build

    We develop and integrate in stages, with deliveries the operation starts using early.

  4. 04

    Support

    We measure, adjust and grow the system alongside the business, in long-term relationships.

One company. One source of truth.

What we do

From diagnosis to a system up and running.

  1. 01

    Consulting and systems architecture →

    We understand how the company operates before we build. We survey processes and information, and design the architecture that defines how systems, data and users relate to each other.

  2. 02

    Custom software development →

    Management platforms that bring together sales, purchasing, inventory, invoicing, accounting and logistics, with permissions, validations, audit trails and traceability.

  3. 03

    Legacy systems modernization →

    We keep what works, fix its limits and guide the transition to modern technology without stopping the daily operation.

  4. 04

    Business intelligence →

    Data models, indicators and dashboards designed from the system up, so management runs on information that is consistent and on time.

  5. 05

    Artificial intelligence and automation →

    Predictive models, intelligent document processing and task automation — when there is a well-defined problem and the data to support it. If the information is confidential, the models run inside the organization.

  6. 06

    Information security and confidentiality →

    We design every system to protect sensitive information: credentials kept in encrypted vaults, least-privilege access for every user and every service, and an audit trail. Confidentiality is part of the architecture, not an add-on.

  7. 07

    Implementation and ongoing support →

    From the initial survey to implementation and continuous improvement: infrastructure, backups and support, in long-term relationships.

What we build with

  • Python
  • Laravel
  • PostgreSQL
  • SQL Server
  • Redis
  • Docker
  • HashiCorp Vault

Case studies

Systems already in operation.

Medical technology · Argentina

Surgical traceability platform

The challenge

Surgeries, implants, instruments and deliveries ran through separate channels: scattered information and little visibility over the operation.

What we built

A platform that joins the whole journey: surgery scheduling, delivery logistics, dispatch notes, consumption and individual traceability of every component, with a sales CRM and visit tracking.

Today

Every department works on the same information and has its indicators up to date.

  • Management system
  • Traceability
  • Logistics
  • Sales CRM
  • Visit tracking

OXXO · Chile and Colombia

Data and stock control

Database setup, migration of the existing information and construction of the stock control system.

  • Databases
  • Migration
  • Inventory

OPC · Argentina

Video surveillance control

Camera control and monitoring system for OPC, a private security company with more than three decades in business.

  • Monitoring
  • Integration
  • 24/7 operation

Analytics · United States

The Signal Boost

Editorial platform for NYU Stern's analytics community: stories of data put to work, told by the people behind them.

  • Web platform
  • Content
  • Analytics

Law firm · Argentina

Private artificial intelligence

The challenge

Professional secrecy ruled out external AI tools: no client document may leave the firm.

What we built

A language model that runs on the firm's own server, tuned to its practice and with its document archive as context. Each lawyer drags documents into a folder on their computer and the model drafts from them.

Today

The lawyers draft with artificial intelligence, and client information never leaves the firm.

  • On-premise AI
  • Confidentiality
  • Assisted drafting
  • Document archive

Our own platform · Argentina

ARCA integration engine

Electronic invoicing and tax controls inside the same management system, connected to ARCA, Argentina's tax authority: no external platforms and no double entry.

500+ million

Argentine pesos invoiced every month

  • Electronic invoicing

    Issuing and authorizing invoices.

  • Invoice verification

    Validity of every invoice received.

  • Taxpayer lookup

    Tax status and details of customers and suppliers.

  • APOC database

    Flags suppliers listed in the fraudulent-invoicing database.

Each company's tax certificates are stored encrypted in a HashiCorp Vault: the system only accesses what it needs to invoice.

  • Our own product
  • ARCA web services
  • HashiCorp Vault

Analytics lab

Applied research.

Our own developments at an experimental stage, built on public data or the literature. They show the methods we apply in real projects.

Joint development · Credit risk

Anticipating defaults on commercial loans

An automated pipeline extracts and normalizes public SEC filings; an XGBoost model estimates the probability of default at 3, 6 and 12 months, and an interactive dashboard with a risk calculator presents the results.

monthly observations
766,558
monthly observations
loans analyzed
24,771
loans analyzed
model variables
205
model variables
of defaults, within the riskiest 10%
36%
of defaults, within the riskiest 10%

Experimental build · Health

Infection risk in spine surgery

Estimates the risk of surgical site infection from the patient's profile and the type of approach. It combines risk factors documented in the medical literature and simulates 10,000 scenarios to return an indicative risk range rather than a single number. It does not replace medical judgment.

risk factors
17
risk factors
surgical approaches
7
surgical approaches
simulated scenarios
10,000
simulated scenarios
risk range of the example (P5–P95)
2.9–5.1%
risk range of the example (P5–P95)

Proof of concept · Consumer goods

Demand and promotions

Demand models with price, seasonality and cannibalization effects across five supermarket brands, and a 13-week promotion plan that maximizes profit with a cap of four promotions per product.

out-of-sample forecast error (down from 89%)
11%
out-of-sample forecast error (down from 89%)
sales of the brand at a 30% discount
×3.4
sales of the brand at a 30% discount
brands modeled, with cannibalization between them
5
brands modeled, with cannibalization between them
weeks of optimized promotions
13
weeks of optimized promotions

Proof of concept · Retail

Inventory policy

Monte Carlo simulation of replenishment policies for a beverage retailer facing uncertain demand, balancing the cost of holding stock against the cost of running out.

average cost against the current policy
−49%
average cost against the current policy
replenishment policies evaluated
6,561
replenishment policies evaluated
runs of 2,000 weeks per policy
100
runs of 2,000 weeks per policy
service level that minimizes cost
67%
service level that minimizes cost

Proof of concept · Subscriptions

User conversion

Logistic regression, random forest and XGBoost models to anticipate which free users of a music subscription service will convert to paid, and pick the thousand most likely for the campaign.

AUC on users the model had never seen
0.83
AUC on users the model had never seen
of those who converted to paid, detected
76%
of those who converted to paid, detected
users in the final test
26,803
users in the final test
users prioritized for the campaign
1,000
users prioritized for the campaign

Our vision

Bringing together technology, information and business knowledge to solve each organization's complex problems.

[email protected]