When The Cloud helps predict the weather
This is the full script of a talk I gave at an AWS public-sector event last year. I’m publishing it unchanged because the ideas still hold: AI works best when it’s treated as a system of learning, interpretation, and human judgment — not magic.
At MetService, our mission is simple: to help everyone be ready for any weather.
We give people the intelligence they need — to plan a weekend football game, land a plane safely, or prepare when severe weather threatens.
Weather intelligence is about understanding change.
We don’t just react to weather — we observe it, we interpret it, we forecast it, we act on it.
That’s how we turn change into understanding, and understanding into action.
What if AI is no different? Just another one of these moments of change we can approach with the same mindset: to learn from it, to make sense of it, and to put it to work.
Learning
When we talk about learning, we often focus on tools and services — and that’s necessary.
But there’s something more fundamental: learning starts with curiosity and grows through imagination.
If we look to the past, we see that technology alone doesn’t drive change.
The real impact comes from people.
History gives us some beautiful examples of that.
When the vinyl record was invented, its creators imagined people using it to send “voice letters” through the mail.
But once people began using it, everything changed. It became music distribution, DJ culture, entire industries.
The real value came when people used it in new ways.
The same happened with the telephone.
In the beginning, Bell used the telephone to send music — a marvel for its time. But people didn’t stop there.
They re-imagined commerce, emergency services, entire new ways of working.
Here’s the pattern: innovation begins with an inventor’s idea, but it matures when users take ownership and use their imagination to transform their world.
Then learning isn’t a destination — it’s a continuous path, fueled by curiosity and anchored to our mission.
We have to learn as much as possible. Unlearn when necessary, so we can learn again.
Interpretation
Once we have enough knowledge, the next challenge is interpretation — how do we make sense of it?
As you get familiar with Amazon Q Developer, you see it can generate code and explain complex systems in plain English.
So, what if we use Q Developer to unlock the knowledge embedded in our legacy systems?
Legacy isn’t just old code — it’s risk, it’s backlog, it’s the barrier between us and innovation.
Now, Are we ready to use this responsibly?
“GenAI hallucinates,” you may say. That’s real. And it’s why we need to work differently.
You can reduce hallucinations drastically if you:
- Augment the model with your own repository as ground truth
- Work in small, reviewable chunks so humans validate every decision
Keeping humans in the loop is essential. There’s no other way to make informed decisions.
“Giving AI access to my whole environment doesn’t feel safe,” you might argue.
Good. That awareness is exactly what responsible engineering looks like.
If direct access feels like too much exposure, add boundaries. Run Q Developer in a Docker Container.
A Docker container gives you isolation — you control what goes in, what comes out, and what it can touch.
That’s the technical boundary. The organizational one matters just as much.
The very first thing we did on my team before using AI was to study and understand the company’s AI Policy. It was clear, detailed, and responsible — outlining exactly how to use the technology safely.
That’s how it should be: curiosity paired with accountability. Innovation grounded in trust.
Execution
At MetService, leadership gave us both the freedom and the context to explore safely.
We didn’t need a huge budget — a few workshops and the free tier of Amazon Q Developer were more than enough to start.
Once we were ready, a two-day hackathon brought it all together.
Our team built a Minimum Viable Product that validated the enterprise value of AI.
Through that process, I came to see AI as three things: speed, accessibility, and empowerment.
It lets us move faster without sacrificing thoughtfulness.
It lowers the barriers to complex work. And it opens the door for more people to contribute to innovation.
Something remarkable happened once the new system reached feature parity: our users immediately asked for new features.
Remove the constraint, and people naturally innovate.
“Think beyond constraints,” encourages AWS Executive Phil LeBrun. And that’s exactly what we saw unfold.
During an internal hackathon, we used Amazon Q Developer to retire a legacy application — 100,000 lines of code — in just two days.
Now we’re free to focus on innovation and impact rather than maintaining legacy systems.
Closing
So I want to ask you: What if your constraints didn’t restrict you anymore?
Imagine what that could mean for your mission, your team, and the people you serve.
It takes curiosity, courage to experiment, and willingness to try.
Because the story of AI won’t be written by its inventors. It will be written by people like us — if we have the courage to take ownership and push beyond what seems possible.
When I see this potential — what we can accomplish — I feel something very powerful: hope.
Hope in our people. Hope in what we can build. Hope in how we can serve.
What amazing tools AWS gave us to fulfill our purpose.
What an amazing time to make innovation a reality.
Truly, what an amazing time to be alive.