Telegram bot · Caja Cercana
Development and launch (2025) of a Telegram bot for ISP technicians that finds the nearest network box to their location. Used internally at the company. Python, SQLite, pytest.
Repository: to be confirmed
I build AI systems that explain themselves and can be audited.
7 years in fiber optics
Remote · UTC-5 · Spanish native · English B1
Code
$ cat ruleset.v1.json
{ "id": "R-02",
"name": "intentos-en-rafaga",
"action": "Limit", "limitMinutes": 15,
"condition": { "type": "burst",
"min": 10, "maxWindowSeconds": 60 } }
$ audit static-example
rule R-02 v1.0.0
decision Limit
used account, failed attempts,
time window
not used name, nationality, language,
device type, gender, source IP
review PendingReviewReal data from the demo: the public ruleset and the static audit example.
Work in production
Telegram bot for ISP technicians. A children's football championships site for a client.
opens a new page
Open Projects →Live demo
An automated decision, with the rule used, the explanation and its audit.
opens a new page
Open the demo →From fiber to AI
Seven years in fiber-optic networks, where things have to work in the field.
opens a new page
Open Journey →How I work
People
- Tests first, then code.
- Nothing is assumed without evidence: an output, a passing test.
- Adversarial review of security, tests and correctness before publishing.
- Real data only. What I do not have is marked as not yet confirmed; never invented figures or achievements.
- I authorize commits and publishing. No agent publishes on its own.
- Every decision is written down: date, alternatives and reason.
- No AI model call per visitor click.
Machines
# Working rules (llms.txt format) - Tests first, then code. - Nothing is assumed without evidence: an output, a passing test. - Adversarial review of security, tests and correctness before publishing. - Real data only. What I do not have is marked as not yet confirmed; never invented figures or achievements. - I authorize commits and publishing. No agent publishes on its own. - Every decision is written down: date, alternatives and reason. - No AI model call per visitor click.
Interest
Sports: jiu-jitsu
I have also done triathlon and obstacle courses (cycling, swimming and track), a few times, not consistently.
What if, in a few years, we could connect the brain to AI and strengthen what is human? I do not know if it will be possible. I think so. (An idea, not a capability.)
Live demo: an automated decision, explained
One click simulates a suspicious access. The system decides, explains why, shows the rule it used, keeps a full audit trail and lets you take control.
Try the demoThe software factory with agents
The way I build: an orchestrator splits the work into delegated tasks with limits, each result passes an automatic review, and a person verifies what the agents report.
- OrchestratorClaude Code orchestrator running on an Ubuntu laptop.
- Delegated tasksEach task has explicit limits: what it may touch and what it may not.
- Automatic reviewEvery result is reviewed automatically before it is accepted.
- Tunnel and accessRemote access goes through a Cloudflare tunnel with access control.
- Human verificationA person verifies what the agents report. Reports are not taken as proof.
Public evidence (repository, captures) is being prepared.
How I work with AI
Using AI is not the point. Governing it is. These are the working rules, short and checkable.
- Tests first, then code.
- Nothing is assumed without evidence: an output, a passing test.
- Adversarial review of security, tests and correctness before publishing.
- Real data only. What I do not have is marked as not yet confirmed; never invented figures or achievements.
- I authorize commits and publishing. No agent publishes on its own.
- Every decision is written down: date, alternatives and reason.
- No AI model call per visitor click.
This site was designed spec-first. The design, plan and decision log are written down before the code.
Demonstrated work
Each item is marked in production or in development, nothing else.
Super Gol Kids
Platform for children's football championships. Public website for a client; I developed the site.
Naunturia
Project in development.
Telegram bot · Box registry
Registers network boxes from the field, with location and evidence, to build a more reliable technical inventory.
Telegram bot · Telecom reports
Technical telecom reports from Telegram: incidents, inspections and follow-up.
Accounting system · React + Flask
Web accounting system with financial dashboard, clients, suppliers, inventory, invoices, payroll and journal entries. MySQL.
Football Manager
Web app to manage matches, teams and players, with filters, profiles and statistics.
HARDLOCK · Android/Kotlin
Android project in Kotlin; its scope is still to be documented.
Skills
Grouped by what I can do. "Shown" means there is proof above; "declared" means I use it but have no public proof yet.
Bots and automation
Shown
Declared
AI systems
Declared
Data and back end
Shown
Declared
Operations
Declared
Journey
- 7 years in the field working with fiber optics.
- There I ran into problems that software could solve.
- I learned Python to solve them myself.
- Today I build AI systems that explain themselves and can be audited.
Criden
For businesses and websites: Criden. A separate product from this site. Link coming soon.
Contact
- Email: [email protected]
- LinkedIn · GitHub
- Download CV (PDF)
- I look for work with artificial intelligence: designing and building systems under human control.
- Long-term interest: how AI relates to the human brain. An interest, not a demonstrated skill.
- Remote, from anywhere in the world. · UTC-5
- Spanish: native. English: B1.
- to be confirmed