Useful AI starts before the chatbot

Reporting automation · Reliable data · Analytics · AI

I automate reports and untangle scattered data.

So teams stop depending on endless Excel, manual Power BI and repetitive processes.

I help you find where time is lost, which numbers don't add up and what to automate first — without building a giant system.

Automated reportsUnified dataTrustworthy metricsAI-ready data

I get into the data mess

The visible problem is rarely the real problem.

I come from support, networking and production applications. That helps me investigate live systems: I do not only look at the report or the chatbot, I look at sources, processes, infrastructure, costs, ownership and how the team operates.

Investigation map

01

Sources

Databases, APIs, files

02

Processes

Loads, rules, checks

03

Model

Metrics and analytics layer

04

Consumption

Dashboards, AI, automation

The goal is not to blame the visible tool. It is to find where the flow breaks.

When I can help

  • You want to use AI or automate on internal data, but the sources are scattered.
  • The process depends on Excel, manual steps or knowledge that lives with one person.
  • People do not fully trust the numbers, or costs and errors appear without explanation.

Target outcome

Move from scattered data to a trustworthy, operable layer ready for analytics, automation or AI.

Services

I can help you if…

  • You build reports by hand every week or every month.
  • Your data is spread across Excel, Power BI, internal systems, databases, CSVs or APIs.
  • Your dashboards are slow, fragile or nobody fully trusts the numbers.
  • You copy and paste information between systems to reach a result.
  • You want to use AI, but first you need to sort out sources, business rules and metrics.
  • You have repetitive admin or operational processes that could be automated.

Entry offer

Reporting and data diagnostic

I review your current process, the data sources, the existing reports and the manual steps that today eat time or cause errors.

You walk away with a clear map:

  • what is happening today
  • where time is lost
  • which numbers don't add up
  • what to automate first
  • what is better left as is
  • a concrete MVP proposal if it makes sense to move forward

You don't need to know whether you need an ETL, an analytical database, a dashboard, AI or an automation. Having a process that is slow, manual or hard to maintain is enough.

Request a diagnostic

Working principle

Most data problems are not where they appear.

A slow dashboard, an unconvincing chatbot, an unexpected cloud bill or a manual reporting process is usually a symptom. The real cause often lives in another layer of the system.

Visible problem → Likely real cause
Unreliable chatbot
→
Unmodeled data, undefined metrics or context scattered across sources
Slow dashboard
→
Analytical model, heavy queries or a warehouse not designed for consumption
High AWS bill
→
Network path, invisible traffic or a service using the wrong route
Manual reporting
→
Process design issue, duplicated logic or unclear ownership

The work starts when we stop looking only at the visible tool and trace the full flow: data, transformations, infrastructure, ownership and consumption.

Data engineering for operational AI

Before adding AI, fix the data.

Most internal AI projects do not fail because of the model. They fail because of what is underneath: scattered data, undefined metrics and business rules nobody ever wrote down.

What is usually broken before the chatbot

  • Scattered sources
  • Fragile ETLs
  • Undefined metrics
  • Hidden business rules
  • Sensitive data without control
  • A chatbot answering over unreliable data

AI does not fix badly modeled data. It only makes it look more accessible.

Real cases

Investigation 01

From scattered data to a layer an AI can query

Symptom: The ask was "we want a chatbot over our data". The real problem was scattered sources, undefined metrics and hidden business rules.

Investigation: Unified relational, document and API sources into ETLs feeding a curated model, defined the metrics and left a queryable layer with limits and control over sensitive data.

Result: A trustworthy base that enables dashboards, automation and AI-assisted queries over data you can actually rely on.

  • Data Engineering
  • ETL
  • PostgreSQL
  • AI
  • Metrics
Read investigation →

Investigation 02

From Power BI bottlenecks to a warehouse-first architecture

Symptom: The visible issue was slow refreshes. The real issue was transformation ownership living inside the BI layer.

Investigation: Traced source queries, Power Query transformations and gateway constraints before moving business logic into Python ETL + PostgreSQL models.

Result: Power BI became a consumption layer instead of the place where core analytical logic lived.

  • Warehouse-first
  • PostgreSQL
  • Power BI
  • ETL
  • Python
Read investigation →

Investigation 03

Why a NAT Gateway suddenly processed ~10 TB/day

Symptom: A cloud cost spike looked like an application issue, but the root cause was a routing path between Loki and S3.

Investigation: Correlated CloudWatch NAT traffic with pod-level Prometheus metrics, isolated Loki and validated the S3 route.

Result: An S3 Gateway Endpoint removed the expensive path and turned the investigation into a reusable FinOps pattern.

  • AWS
  • FinOps
  • Kubernetes
  • Loki
  • Networking
Read investigation →

Investigation 04

Low-cost static hosting architecture for a real business

Symptom: A traditional business needed a professional web presence without adding operational weight or monthly platform cost.

Investigation: Used static hosting with S3, Cloudflare and Terraform, keeping the origin restricted and the delivery path simple.

Result: Fast, secure and maintainable hosting with practically zero operating cost.

  • AWS S3
  • Cloudflare
  • Terraform
  • Static Hosting
  • Security
Read investigation →

How I think

How I investigate systems

I treat data problems like system incidents: first separate symptom from cause, then trace the flow until the reliability failure becomes visible.

  1. 01

    Visible symptom

    What business sees: slow dashboard, unreliable chatbot, inconsistent KPIs, high costs or unstable refreshes.

  2. 02

    Trace the flow

    Sources, transformation, infrastructure and consumption. Data changes shape at every layer.

  3. 03

    Find operational friction

    Manual processes, duplicated logic, unclear ownership, expensive routes or weak validation.

  4. 04

    Reduce complexity

    Solve with simplification, automation or a targeted redesign. Do not add complexity by reflex.

About

I'm Luis Iván Payero, but most people call me Luigi. I come from support, networking and production incidents. That changed the way I work with data.

I learned to look at live systems: what changed, what depends on what, where the flow breaks and who needs to operate the solution later. That experience matters when a report fails, a pipeline breaks or a cloud cost appears without explanation.

My focus is not only building pipelines. It is understanding why a system stopped being reliable and leaving behind a clearer, maintainable and operable architecture: a base ready for analytics, automation or AI, not a rushed chatbot over unsorted data.

Luis Iván Payero

Technical stack

AWS, PostgreSQL, Python, Airflow and Power BI to build data layers ready for analytics, automation or AI.

I do not use tools as identity. I use them when they reduce risk, clarify ownership or make the system more operable. AI comes later, on top of a trustworthy base.

  • AWS
  • PostgreSQL
  • Python
  • Airflow
  • Power BI
  • SQL
  • ETL
  • FinOps
See cases

Real delivery

Other digital projects

Small projects where the value was shipping something simple, clear and useful. This is not the center of the site; it only shows the ability to turn a real need into production.

Casa Lucho — professional painter

Problem: A painter needed a professional web presence to show work and capture inquiries without adding operational cost.

Solution: Static architecture with S3, Cloudflare and Terraform.

Result: Fast, easy to maintain and practically zero-cost hosting.

View live site →

Diego Rago — pixel artist

Problem: A pixel artist needed a portfolio to showcase work and style with strong visual focus.

Solution: Lightweight landing page with a clean gallery and clear contact.

Result: An online presence focused on the work itself, no noise around it.

View live site →

Richard Scobar — tattoo studio

Problem: The studio needed to explain style, work and contact clearly.

Solution: Direct structure, visual content and a simple CTA.

Result: Less friction for real inquiries.

View live site →

Next step

Send me the report, process or problem that's slowing you down today.

You don't need to have the solution figured out, or a formal brief. Just tell me which process is eating your time or which number doesn't add up — that's enough to start.

Tell me about your case Message me on LinkedIn