Service lines

How I can help

Four ways I can help, depending on your situation. Each one says what it includes and what you get at the end. If you are not sure which fits, message me and I will tell you.

If you want AI doing real work in your business

AI integration into your business

I connect AI models to the information and systems you already use, so they answer from your data, read your documents and do real work. Every integration has clear permissions, a record of what it does and tests that measure how well it answers.

What's included

Assistants that answer by text or voice based on your documents and information, cite their source and say so when they cannot find the answer.
AI inside your current systems: it classifies, summarizes, extracts data from documents and generates reports in seconds.
Agents that look things up and take actions in your systems, from the web or WhatsApp, only with the permissions you set and with your team supervising.

Concrete deliverable

AI running inside your operation, connected to your data, with tests that show how well it answers.

See case: SIMDA See case: IUK English

Technologies (for your technical team)

LLMsRAGLangGraphFastAPIPostgreSQLEvaluations

If you need your own system, not a generic tool

Custom software and web development

I design and build complete software: frontend, backend, database and deployment, all in one pair of hands. My platforms are already in production: IUK English's learning platform shipped in 3 weeks and the Mexican Navy's in 8.

What's included

Complete web platforms with users, permissions, an admin panel and integrations with your other systems, plus AI where it adds value.
Internal systems and dashboards that bring your data into a single view, with maps, alerts and reports to decide.
Fast, bilingual, accessible websites optimized for search engines, easy to update.

Concrete deliverable

Software in production with your real users and data, documented code and manuals to operate it.

See case: IUK English See case: SIMDA

Technologies (for your technical team)

Next.jsReactTypeScriptAstroFastAPINode.jsPostgreSQLDocker

If you lose hours copying data or reviewing PDFs

Automation of manual work

I turn what your team does by hand (copying from spreadsheets, reviewing PDFs, entering data) into processes that run on their own and leave the information ready to use.

What's included

Your Excel files and Google Sheets, cleaned and validated in a database, ready for dashboards, applications or agents.
Scheduled processes that run on their own: they sync data, check sources and alert you when something does not add up.
Data extracted from PDFs and scans, each with its source and a confidence level so you can verify it quickly.

Concrete deliverable

Automated processes running on your real files, with the data ready wherever your team needs it.

Technologies (for your technical team)

PythonPandasGoogle Sheets APIPostgreSQLOCRLLMs

If you already have a system that underperforms

Rescue, acceleration and training

I thoroughly review AI systems, automations or computations that do not work, are slow or cost more than they should. I find the cause, fix it and, if you need it, train your team so it can maintain the system on its own.

What's included

A diagnosis of where money or time is being wasted, with a clear plan of what to fix first and the code already corrected.
Computations and simulations that take days, redesigned to run in hours on multiple cores or on GPU.
Training on your company's real code: how to build, measure and maintain AI systems without depending on third parties.

Concrete deliverable

The system fixed, with before-and-after measured improvements, and your team ready to maintain it.

Technologies (for your technical team)

ProfilingCUDAOpenMPDockerPython / C++Evaluations

Option for any service: AI on your own servers. Everything can run inside your infrastructure, with no data leaving your organization, as in the Mexican Navy platform. Or in the cloud, if you prefer speed and lower cost.

Ways to work together

Clear formats, no surprises and no strings attached

First step

Free quote

You tell me what you want to build and I send you scope, estimated time and price. For small projects it takes a few days; for large ones, we first define the scope together.

Defined scope

Milestone-based delivery

The first version reaches production in 3 to 8 weeks depending on scope, and then we keep evolving it in stages. At each stage you see and test something working, not just a promise.

Continuity

Managed operation and support

I keep an eye on the system, update the security controls and give you technical support when you need it.

Methodology

How I build it: measured at every step

A measurement-driven engineering process: before building, we agree on what working well means and check it at every phase, with clear dates, auditable code and tangible deliverables. The first version reaches production in 3 to 8 weeks depending on scope, and then keeps improving with what is learned in use.

01 · At the start
Discovery and success metrics
I learn your process and we agree on measurable success metrics and a dated plan.
02 · Before building
Test cases and baseline
I gather real cases from your operation, with their correct answer, and measure how they are handled today.
03 · First weeks
Prototype and iteration
I start with the simplest solution and only add complexity when the tests justify it.
04 · Before launch
Security and robustness
Permissions, data protection, abuse testing, human fallback and cost limits.
05 · Launch
Gradual launch
First with a pilot group, then with everyone, with monitoring and training.
06 · Ongoing
Monitoring and continuous improvement
Every real failure becomes a new test, and every improvement is checked before release.

Improvement loop. What is learned in production goes back to phase 02 as new tests.

01At the start

Discovery and success metrics

I thoroughly review your process, your data and your goals, and we define together what working well means: how accurate, how fast and how much each operation can cost. I also flag risks and tell you whether AI is the right fit or a simpler solution is.

Deliverable

A plan with dates, deliverables and target metrics, approved by both of us before starting.

02Before building

Test cases and baseline

I gather real examples from your operation with the expected correct answer, including the hard cases and the ones that cause errors today. I measure how they are handled now to get an objective baseline, so every step forward is shown with numbers, not impressions.

Deliverable

A test set built from your operation and a measurement of how those cases are handled today.

03First weeks

Prototype and iteration

I build the simplest version that could work and measure it against the tests. I review the failures one by one, adjust and measure again. I only add search over your documents, agents or custom models when the results justify it, and I compare providers to cut cost.

Deliverable

A version you can try yourself, and how each iteration compares against the agreed metrics.

04Before launch

Security and robustness

I harden the system: clear permissions on who can see and do what, protection of sensitive data, tests so nobody can trick it with hidden instructions, a human fallback when the AI is unsure, cost limits and tests under heavy simultaneous use.

Deliverable

The complete system in a test environment, with the security review documented.

05Launch

Gradual launch

I put it live on your servers or in the cloud, first with a pilot group and then with all users. Every operation is logged so problems are caught early, and I train your team to run it with confidence.

Deliverable

The system in production with real users, monitoring dashboards and clear manuals to run it.

06Ongoing

Monitoring and continuous improvement

With the system in use, I follow answer quality, cost and speed, and review the cases where it fails. Every failure is added to the tests, and every improvement (or provider model change) is checked against all of them before release, so it gets better without breaking what already works.What is learned in production goes back to phase 02 as new tests.

Deliverable

New versions checked against the full test set before they reach your users.