Two fields, one craft.
OSS/BSS and applied AI look different, but they solve the same problem: turning business rules into reliable systems. This map shows how they connect.
OSS/BSS
Where it all started.
OSS/BSS are the systems a telecom operator uses to sell, deliver and bill its services. For 25 years I've turned operators' business rules into these systems, for mobile and fixed operators across Latin America.
From catalog to bill
Select a stage in an order's journey.
The AI layer doesn't replace any system: it connects to them. Flow Audit is under construction.
What I've done
Designed catalogs on TM Forum SID. One was rationalized from more than 2,000 offers to just over 500.
Where AI comes in
Knowman searches the documentation by meaning and answers with its sources.
What I've done
Integrated the catalog with order management and the CRM.
Where AI comes in
Flow Audit, under construction, follows the order through this stage to show where it gets stuck, without replacing any system.
What I've done
Built SIP+, a BPM engine, and implemented order orchestration with Oracle OSM.
Where AI comes in
Flow Audit, under construction, follows the order through this stage to show where it gets stuck, without replacing any system.
What I've done
Worked on network management tools and their migration between vendors.
Where AI comes in
Flow Audit, under construction, follows the order through this stage to show where it gets stuck, without replacing any system.
What I've done
Integrated the catalog with Charging System and BSCS, and led the catalog workstream in the migration to DMP 22.
Where AI comes in
Sentinel Engine decides on charge disputes with rules in code and customer confirmation.
ECM · CBiO · DMP 22 · Order Care · Charging System · BSCS · Oracle OSM · TM Forum SID · Salesforce · SAFe
Flow Audit, my project under construction, builds on this experience.
Applied AI
The same discipline, with new tools.
AI is useful when it fits the business. I combine language models with rules, search and evaluation.
How an agent decides
This is how Sentinel Engine handles a request. Select each step.
The customer writes in their own words, in Spanish or Portuguese.
The model only identifies what the customer wants. It doesn't decide anything.
The bank's policy, data protection and validations live in code, not in the prompt.
Nothing runs until the customer confirms. When needed, the case goes to an advisor.
Every decision is recorded, and every version passes an evaluation before release.
RAG and semantic search
Search by meaning, not just keywords, with answers that cite their sources. Knowman uses Postgres with pgvector and local models with Ollama.
Evaluation and testing
Every prompt and every change is measured against reference data before release, and the results are saved so they can be compared.
Continuous learning
Learning has always been part of the job.
These are courses and certifications, not production experience. They show my interests and where my work is heading.
- 2026
Agentic AI (Skillsoft): foundations, architectures, LangChain, LangGraph and RAG.
- 2025
Generative AI Driver's License, machine learning fundamentals and automation (Ericsson).
- 2023
Docker and Kubernetes; SAFe 6 Product Owner and Scrum Master.
- 2020 – 2021
Cloud fundamentals (OpenStack, containers, NFV) and 5G (Ericsson Academy); SAFe 5.
- 2014 – 2015
Oracle Communications OSM: sales, pre-sales and implementation.
Clients and teams
Systems are built with people. A large part of my work has been listening to clients, agreeing on designs and delivering with my team.
Requirements and sign-off
Day-to-day work with operators' commercial, product and operations teams, from the business need to an approved design.
Offers and proposals
Commercial offer design, technical scope and effort estimates for B2B proposals.
Leading teams
Teams of 4 to 8 engineers from design to delivery, with coaching. Certified SAFe Product Owner and Scrum Master.
Communication
Designs presented in Spanish to clients and in English to global teams. Training in communication and storytelling.