In geometry, a locus is the set of all points that satisfy a condition. Our work is to find that condition in your business or system: to separate signal from noise until it takes a shape you can read and act on.
Ask your business.
An AI agent for BI and analytics, connected to your sales, finance, marketing and logistics systems. It reads your transactions and events and answers in plain language, with numbers, charts and recommendations.
Fig. 1 Logistics agent: route network, hub cities and profitability by route. Wait times sped up. Demo in Spanish; English version coming soon.Request a session
Three layers of one system.
AI isn't installed. It's built on data you can trust. Without an operational foundation, an agent just repeats the mess, faster.
First, we map where you stand.
Many organizations still run on spreadsheets and systems that don't talk to each other. An organization, like a body, needs to sense its own state and its position in the world before it can move with precision. So we start by mapping where your operation stands: against your own data, against your competitors and against the AI moment. From there we move forward, one layer at a time.
Stage 0Spreadsheets and isolated systems
Stage 1Data foundation
Stage 2Reliable reporting
Stage 3Prediction and agents
Stage 1
Data engineering
The foundation. We connect your systems and model them in the cloud or in a hybrid setup, designing for cost from day one. Data that is complete, consistent, traceable, observable and on time.
Solves
Spreadsheets and systems that don't talk to each other
Delivers
Production pipelines, a modeled warehouse, validation and documentation
The reading. Executive dashboards, marketing attribution and reports leadership trusts, because every number can be traced back to its source.
Solves
Reports that don't match across teams
Delivers
Dashboards, attribution models, query agents
With
Power BI, Looker, Plotly, ggplot
Fig. 3 A surface emerges from the dots.
Stage 3
Data science and AI
The intelligence. Forecasts, models and agents that work on top of your foundation, not in spite of it. AI comes in at four points:
Building
Pipelines written with AI and reviewed by people
Extraction
Documents, images and scans turned into structured data
RAG
Answers grounded in your own content, cataloged and current
Interface
Plain-language questions alongside your dashboards, not instead of them
With
Python, R, PySpark, XGBoost, LangChain, embeddings, vector databases, OpenAI and Claude
Fig. 4 The relief of information.
How we build a locus.
Four steps, always in this order.
Step 1 of 4
Define the condition
We start by pinning down, with you, the decision you want to make and what would have to be true to make it well. The question comes before the data.
Step 2 of 4
Gather the points
We connect and organize the sources that matter: sales, operations, marketing, customers. Only the ones you need.
Step 3 of 4
Reveal the shape
We analyze and model until the pattern appears: the points that meet the condition draw a shape you can read.
Step 4 of 4
Action
We put the shape to work: reports, dashboards, assistants alongside the dashboard, agents and models your team uses every day to make better decisions.
What it looks like inside.
We show you the architecture before we build it, so you know exactly what you're getting.
Fig. 5 Reference architecture. Every client gets their own.
For whoever makes the call.
Executive leadership
Where are we winning, and where are we losing?
Sales and eCommerce
Which products will run out, and in which region?
Finance
What happens to cash flow next quarter?
Marketing
Which campaign brought customers, and which only brought clicks?
Logistics
Which routes and loads are costing us too much in fuel?
Government and nonprofits
Where is the budget actually going?
What we've solved.
More than nine years building data infrastructure for eCommerce, SaaS, AgTech, real estate and healthcare. Based in Puebla and Mexico City, we work nearshore, in your time zone, with clients across the U.S. and Mexico.
Multi-brand eCommerce
Inventory spread across several accounts and warehouses, with no unified view; reconciliation was manual.
A single view of inventory. Manual reconciliation became a query that runs in seconds.
Online store
Tens of thousands of orders silently marked as undelivered because of a platform limitation.
Root cause diagnosed and fixed in three lines, without touching the rest of the pipeline.
B2B SaaS
CRM attribution counted the same revenue across several campaigns and inflated the pipeline.
A model that credits a single campaign per opportunity, now part of the basis for budget planning.
Marketing agencies
Marketing calls and form fills with no link to the jobs they sold.
End-to-end attribution in production, running in seconds.
Automated vertical farm in the U.S.
Needed data modeling, ingestion, analytics and data science.
Full-stack development and statistical models that isolate which environmental variables drive crop yield.
Real estate
Property prices set by gut feeling.
An ML model that predicts home prices from online listing data.
Healthcare
Quality metrics with no visibility by physician or location.
Role-based compliance dashboards and monthly and weekly patient readmission forecasts.
Fig. 6 Actual vs. predicted price. Data collected through web scraping, processed in an MLOps pipeline and modeled with XGBoost. Illustrative data, calibrated to an R² of about 0.70.
Who we are.
Good decisions start with knowing where you stand. Every technology cycle (electricity, the internet, and now artificial intelligence) follows an S-curve, and the organizations that make the most of it are not the ones that run fastest, but the ones that locate their position on it precisely.
Locus exists for that: to help companies and organizations separate signal from noise in their data and turn it into decisions that serve people. We have been doing it since 2016.
Adrián del Valle Rodríguez
Founder · Data Lead
More than nine years designing end-to-end data infrastructure for eCommerce, B2B SaaS, AgTech and marketing agencies in Mexico and the United States: pipelines, cloud data warehouses, models and the dashboards leadership decides with. A graduate of the UDLAP School of Business and Economics, with training in Data Science and Engineering at UC Berkeley.
Ph.D. in computer vision, with two U.S. patents. He founded Prefixa, a 3D and extended-reality company that had its exit in 2018, and is now co-founder and CEO of Vulcanis, which builds deep-tech sensors for mining and space. At Locus, he advises on the technical direction of projects.
Most likely. We work every day with Shopify, Amazon, MercadoLibre, TikTok Shop, Salesforce, HubSpot, Meta Ads, Google Ads and GA4, and we build custom connectors when none exist.
Do I need my own technical team?
No. We handle the build and leave your team with what it needs to use it. If you already have analysts, we train them and work alongside them.
What happens when the project ends?
Everything is documented: architecture, runbooks and handoff sessions with your team. The system is yours and doesn't depend on us. If you'd like, we stay on with a monthly plan for maintenance, updates and support.
How do you keep my data secure?
Your data lives in the cloud of your choice, or on your own infrastructure, with backups and with access limited to the credentials of the people responsible for the project. We sign an NDA before we touch any data.
How long until a first result?
It depends on your organization's digital maturity and data culture. A pilot typically takes 30 to 40 business days. For AI in production, we set the timeline with you in a tailored session.
Request a session.
We review every request to make sure we can add value. If there's a good fit, we can build a demo with a sample of your own data, under NDA.
