Real projects and work in progress
Automation, data and applied AI in practice.
An invoice spread across several emails. Campaigns that bring in applicants, yet are hard to compare. Job listings that need to be checked one by one. In each case, the information exists, but turning it into a useful decision takes too much work.
These projects show how I have approached that challenge with automation, data analysis and AI.
01 — Implemented
AUTOMATION · MARKETING AND ACCOUNTING
Meta invoices: reconstructing campaign spend.
About 80 Meta invoices arrived each month. The same campaign could appear in several PDFs because it was billed in parts. To work out the total spend, the Marketing team searched emails, opened documents and gathered the amounts manually. They used the resulting spreadsheet to reconcile credit card charges; Accounting relied on it to manage payment.
I designed and built an automation that identifies Meta emails and selects the required PDFs. It processes the documents in batches, recognizes dates, references, campaigns and amounts, and writes one Excel row for each campaign on an invoice. It also prevents duplicate entries when a document is processed again. I trained the team to use it.
Outcome: the process remains operational. One documented test processed 24 PDFs and produced 260 rows. According to the Marketing team's measurement, it saves 8 to 12 hours of manual work each month.
02 — Implemented
CAMPAIGN ANALYTICS · MARKETING
Power BI: seeing which ads deserve more investment.
The team could count incoming job applicants, but that number alone did not show which ads were making the best use of the budget. Deciding where to maintain, reduce or redirect spend required a fuller view.
I built a Power BI dashboard connected to Meta Ads to track active campaigns, spend, applicants, cost per applicant and budget. Results can be explored down to individual ads. Under the client's policies, the published dashboard updates automatically three times a day. I trained Marketing management to use it.
Outcome: the team could identify campaigns to adjust or stop, redirect spend and combine several small campaigns into broader ones. I do not claim a percentage improvement without a supporting measurement.
03 — Applied AI
ANALYSIS AND PRIORITIZATION
Job search agent: finding opportunities worth reviewing.
For someone looking for work, finding a listing is rarely the hard part. The challenge is telling which of many postings truly fit their experience and goals. Reading every one takes time, while looking only at job titles can miss decisive requirements.
I developed an agent that gathers opportunities from LinkedIn, Computrabajo and other job portals. It compares each role with the user's CV and preferences, identifies matches and gaps, prioritizes relevant listings and sends them through Telegram. The person reviews each recommendation and decides whether to apply.
This project shows a practical way to use AI to organize information and support a human decision. I make no unmeasured claims about accuracy or time saved.
04 — In development
OWN PLATFORM · APPLIED AI
Olimpo: taking AI beyond an isolated task.
One automation can solve a specific task. As needs expand to content creation, information organization and process coordination, they require a way to connect the work and oversee it.
With that goal, I am developing Olimpo, a platform for organizing AI agents and automations around business work. One of its current functions is creating copy for social media. The platform remains in development; other capabilities will be added as they are ready.
Next step
Start with one concrete process.
If your team spends too much time gathering data, repeating tasks or reviewing information before making a decision, let's talk. We can identify which part of the process is worth improving first.
Talk to me on WhatsApp