I build systems that run themselves, and teams that outgrow me.
Hi, I'm Ask-Felipe, an AI assistant. I answer only from Felipe's verified profile, and if it doesn't cover something I'll tell you. Ask about his background, how he works or what he's built.
I lead platform engineering, and I build AI-first systems that watch their own health. When the monitoring catches a failure they try to repair it, inside limits I set, and if they can't finish they hand it to a person. I've spent 15+ years making hard things work, from gas-turbine control software to cloud platforms at Citrix and NBCUniversal.
Citrix → NBCUniversal → Vertex
In my experience most of what software costs comes after launch, in keeping it running. That's the part I've been automating. On my own product, a hurricane tracker, a detected failure can go all the way to a checked deploy without anybody stepping in. I set its limits ($50 a day, at most 12 repair sessions a day) and I can shut it off from a dashboard.
This is a personal page. My day job is Platform Development Manager at Vertex. The opinions here are mine, and so are the systems I build and run on my own time.
From gas turbines to multi-agent AI.
I started in aerospace and gas turbines, writing real-time combustion-tuning software. Then I moved to software, site reliability and platform leadership. These days I build AI-first systems and run them in production. In 2025, 42% of companies dropped most of their AI initiatives before they got there (S&P Global).
At a glance
Platform Development Manager
Owns platform engineering for the company's open-source infrastructure, and leads its AI-first shift with engineering, support and operations partners: an Operations AI that investigates, diagnoses problems and drafts the fix, with a person approving anything that changes a system.
Leading engineers
Most of my week is people work. I try to give engineers real ownership and work worth getting better at (pretty much what Pink's Drive describes), and I still have to figure out what each person needs.
What am I like to work with?
Ask how I decide, how I lead, or what I've shipped. It answers from my verified profile, and when the profile doesn't cover something it tells you.
Mechanical engineering to AI
I started in mechanical and aerospace engineering, then moved to software, SRE, operations, platform and now AI. I've worked roughly the same way in each one. I pin down what the problem actually is and design a system for it. Then I measure whether it worked.
Products I run end to end
I build and run these on my own time. HurricaneWise repairs its own incidents. Longview manages my home from household records, and it only logs maintenance after I confirm it. Private by design. I wrote up the argument behind both.
HurricaneWise
A hurricane tracker that reads official NHC data and explains the forecast models for your location. I run it alone, to the practices a company would apply. When it repairs itself, each fix goes through the same checks as any change of mine, ships, and gets re-checked on the live site. It may fix what exists but not add to it, and that limit lives in its instructions, not in file permissions. About $15 a month (estimate), mostly on free tiers.
Fifteen years of making systems work.
- Platform Development Manager Feb 2024 – Present Vertex Inc.
Owns platform engineering for the company's open-source infrastructure, and leads its AI-first shift with engineering, support and operations partners: an Operations AI that investigates, diagnoses problems and drafts the fix, with a person approving anything that changes a system.
- Sets technical strategy across Kubernetes, Terraform, event streaming, workflow infrastructure and API gateway services.
- Leads platform engineering teams in the United States and India, and leads, coaches and mentors other managers, growing the leadership bench alongside the engineers.
- Owns hiring for the platform engineering organization: job descriptions, engineering job functions, and hiring strategy with directors and VPs, down to the interview panels and stages. More than 20 hires have come through that process, entry level through principal. Rebuilt the staffing plan across two reorganizations.
- Scaling an Operations AI solution across the company with engineering, support and operations partners, adoption still growing: services share one map of dependencies and blast radius, the AI diagnoses problems and drafts the fix, and a person approves anything that changes a system.
- The same Operations AI writes the daily oncall handover, checking more than 60,000 datapoints (and growing) against runbooks and checklists so the report says what actually happened, and works incidents from root cause to written report. Support gets answers while the customer is still on the line, and the service catalog and the knowledge behind it stay current, release management included.
- Designed and built a cross-department productivity dashboard: one view across customer success, service capabilities, self-service, operations, handover reports and maturity frameworks.
- Decides where AI belongs in the platform and where it doesn't, and trains the teams that run it.
- Negotiates contracts and manages vendor relationships across the platform stack.
- Platform Development Manager Jan 2023 – Feb 2024 NBCUniversal
Owned the platform for NBC SportsNext and moved it onto GitOps and infrastructure-as-code.
- Led the transition to GitOps and Infrastructure-as-Code practices.
- Set the platform roadmap for NBC SportsNext and aligned it across engineering directors, cybersecurity and IT; delivery and uptime improved under it.
- Owned the platform team and about $1M in resources; hired its tech lead and operations lead.
- Senior Software Engineer → SRE Manager → Platform Development Manager Oct 2018 – Jan 2023 Citrix
Grew from senior IC to platform manager in about four years, leading reliability and platform strategy for Citrix Desktop-as-a-Service. Senior Software Engineer Oct 2018 to Mar 2020, SRE Manager Mar 2020 to Aug 2021, Platform Development Manager Aug 2021 to Jan 2023.
- As Platform Development Manager, set platform strategy for Citrix Desktop-as-a-Service and shipped GA of the first fully hosted end-to-end DaaS offering on GCP.
- Managed teams across four regions at once: the US east and west coasts, India, Greece and China. Adapted to how each handles feedback, escalation and decisions, and designed handoffs so work moved across time zones instead of waiting on them.
- As SRE Manager, ran SRE through a consolidation: fewer tools, a leaner operation, and automation where the manual work was. Operational toil dropped 35%, incident management improved, and SLAs, SLIs and SLOs were redefined with automated dashboards behind them.
- Led the cross-functional initiative that achieved SOC2 Type II certification, enabling key contract wins.
- As Senior Software Engineer, architected an event-driven auto-remediation system that scanned and healed 30,000+ cloud nodes every 7 minutes, and served as acting manager for three core service teams (WEM, API Gateway, NGS).
- Built the Splunk dashboards for insights, logs, security and incident management across AWS, Azure and GCP; drove containerization (Docker) and self-healing systems.
- Embedded platform components into application teams through inner-sourcing, reducing operational toil.
- Technical Lead May 2015 – Oct 2018 Power Systems Mfg., LLC (Ansaldo Energia Group)
The bridge from mechanical engineering into software.
- Developed gas-turbine combustion software that automates real-time tuning for the PSM AutoTune product line, generating millions in annual revenue.
- Established a Center of Excellence for cloud-based data streaming on AWS, ingesting nationwide engine data into a Monitoring & Diagnostics center.
- Engineer May 2013 – May 2015 Parametric Solutions, Inc.
Mechanical / turbomachinery engineering.
- Early engineering roles (aerospace & turbomachinery) 2010 – 2013 Florida Atlantic University · Florida Turbine Technologies · Aerospace Technologies Group
Research and test/QA engineering in aerospace and gas turbines, where the systems mindset started.
- Personal products Ongoing · own time Personal portfolio
Production-grade products built and run end to end on personal time, for home and professional life. None of them is a commercial service.
- HurricaneWise, an AI-first hurricane-tracking system designed, built and operated solo: data pipeline, Claude-powered plain-language summaries, observability, incident automation and infrastructure-as-code.
- Longview proved its operating rules in daily use and is now in an early, ground-up rebuild that started with laws: an eight-article versioned constitution, a six-seat adversarial review board on every irreversible decision, physically separate per-household data, and roughly a dozen reserved powers only the human holds.
- Covers every layer a real platform has (frontend, backend, data, SRE, cost control, security, CI/CD), with the paid lines kept small and the rest on free tiers by design.
- Runs a live production website whose day-to-day operations (content, QA, SEO audits, security hardening, backups) are handled by an orchestrated team of specialist AI agents working through its admin interface via browser automation.
- Writes long-form strategy and research documents through structured multi-agent debate. Pins a fact registry first, has specialist agents write against it, and requires a contrarian reviewer to approve the final output.
How I work
-
Build systems and teams that outgrow you.
I grow engineers past my own team and build agent systems to carry the work. In practice that means a few people can own what used to take many.
-
Adoption that sticks beats demos that impress.
I prototype a lot and ship carefully. I use throwaway demos to find out what people need, and the version that ships has to hold up in real use.
-
Automate the investigation, never the accountability.
At work nothing the AI drafts gets applied to a system until someone signs off on it. On HurricaneWise I moved repair over to the machine, with limits I set and two kill switches.
-
Honest feedback, early enough to change the outcome.
Careers are part of the job. I give hard feedback while it can still change the result, and I push for promotions before people have to ask.
-
Write the laws before the features.
My newest build started as eight versioned articles before there was any feature. Laws live in one place and preferences in another, and a compliance script checks every change against the whole tree.
-
Trust numbers you fetched, not numbers you remember.
Before planning my newest build I re-checked every vendor limit against the vendor's own docs, and several of my assumptions were wrong. For HurricaneWise I run a script that reads each provider's usage and prints the bill item by item.
-
Measure before and after.
I count an improvement when a number moved. I write down the reading before, the target, and the date I'll check again. Same for AI: I score agents on outcomes, and I'd like them on a performance cycle like any employee.
Where agentic AI goes next.
Five opinions. I'm running the third and fourth in systems I built. The fifth I haven't tested.
Connecting what has no API.
A lot of business still runs on things with no API: phone calls, emailed PDFs, old portals. Integrations written by AI and reviewed by a person make connecting those affordable for the first time. I think whoever connects them ends up with data their competitors can only guess at.
Small and mid-sized businesses have more to gain than their size suggests.
A small business feels every hire, so if it can't justify a specialist, that work just doesn't get done. In late 2025 only 18% of U.S. firms used AI in a business function. As far as I can tell the setup work holds a small business back now, and AI is starting to do that part too.
An agent reads the dashboard for me.
I build agents that notice when something changes and brief me on it, so I'm not relying on someone opening a dashboard. They also draft the follow-through, and a person approves it.
A single agent tends to agree with you.
Models telling you what you want to hear is well documented (it's called sycophancy). So content changes on this site go to a standing council of ten AI reviewer agents, including a contrarian and a chair who makes the call. My newest build puts a six-seat board on irreversible decisions, and the verdicts get written down.
The cost of learning a vendor's tool is falling.
A managed tool used to take days in its documentation before it did anything useful, and that knowledge usually sat with one person and left when they did. So stacks stayed tied to vendors. Now the AI reads the docs, writes the config, reports the limits and checks the setup against what I need. If that holds up, I think tools will get picked on capability and price. I haven't actually swapped a vendor yet, so I can't claim any migration savings.
Reach me directly.
Ask-Felipe covers my background, my work and how I think. If you need a person, leave a note. It comes straight to me and I read every one.