First Steps into AI: How Teaching Machines Math Changed My Career Path
Published: September 20, 2025 • 9 min read
After a year of automating workflows and creating training videos at Cambium Inc, I thought I had a pretty good grasp on where technology was heading. Then I stumbled into an opportunity that would completely reshape my understanding of what was possible: training AI models to understand and solve mathematical problems at Outlier.
Little did I know that this experience would become the foundation for everything that followed in my career. Sometimes the most transformative opportunities come disguised as something completely different from what you think you want to do.
Discovering the World of AI Training
The role was straightforward on paper: AI Mathematics Trainer at Outlier. My job was to evaluate and rank AI-generated responses for mathematical accuracy and relevance, generate complex mathematical questions and solutions for AI model training, and conduct detailed reviews of AI outputs to ensure factual accuracy and educational value.
But here's what the job description didn't capture: I was essentially becoming a teacher for artificial intelligence. Every question I crafted, every response I evaluated, every correction I made was helping shape how AI systems would understand and approach mathematical reasoning.
Coming from a Computer Science background, I had solid mathematical foundations, but this role required something different. It wasn't just about knowing the right answers—it was about understanding how to teach those concepts to a completely different kind of intelligence.
The Art and Science of AI Evaluation
Understanding What AI Gets Right (and Wrong)
One of the most fascinating aspects of my work was seeing the patterns in how AI models approached mathematical problems. Sometimes they would nail complex calculus problems but stumble on basic arithmetic. Other times, they'd provide mathematically correct answers but explain them in ways that would confuse students.
My job was to catch these nuances and provide feedback that would help the models become better mathematical tutors, not just mathematical calculators. This required me to think about math education from an entirely new perspective.
Creating "Golden Standard" Questions
Generating complex mathematical questions for AI training wasn't just about making problems difficult—it was about creating scenarios that would challenge the AI in specific ways. I learned to craft questions that would test not just computational ability, but conceptual understanding, step-by-step reasoning, and the ability to explain complex ideas clearly.
Each question I created had to serve multiple purposes: it needed to be mathematically sound, educationally valuable, appropriately challenging, and designed to reveal specific aspects of the AI's reasoning process. It was like being a mathematician, educator, and cognitive scientist all at once.
The Feedback Loop That Changed Everything
The most rewarding part of the role was seeing how my feedback actually improved the AI models over time. When I identified a pattern of errors or highlighted particularly good explanations, those insights would get incorporated into the training process.
I was literally watching artificial intelligence get smarter, and I was part of making that happen. There's something profound about being involved in teaching machines to think—it makes you reflect deeply on how learning actually works.
Skills I Never Expected to Develop
Prompt Engineering Before I Knew What It Was
Without realizing it, I was learning the fundamentals of what would later become known as prompt engineering. When crafting mathematical questions for AI training, I had to be incredibly precise with my language. A small change in wording could completely alter how the AI interpreted and solved a problem.
I learned that the way you ask a question to an AI is just as important as the question itself. This skill—understanding how to communicate effectively with AI systems—would prove invaluable in everything that followed.
Quality Assurance for AI Systems
Conducting systematic reviews of AI outputs taught me to think like a quality assurance specialist for artificial intelligence. I developed an eye for spotting subtle errors, inconsistencies in reasoning, and gaps in explanations that might not be obvious at first glance.
This experience gave me a deep appreciation for the importance of human oversight in AI systems. The models were incredibly powerful, but they needed human judgment to ensure their outputs were not just correct, but genuinely helpful.
Working in a Distributed Research Environment
The role was completely remote and largely independent. I had to manage my own workflow, meet strict quality standards without constant supervision, and collaborate with a distributed team I rarely met face-to-face.
This taught me the discipline and self-direction that would prove essential for all my subsequent remote work. It also showed me how research and development actually happens in the AI field—it's not just brilliant scientists in labs, but a coordinated effort involving many people with different expertise.
The Bigger Picture: Contributing to AI Education
Learning How AI Actually Learns
Working so closely with AI training gave me insights into how these systems actually develop their capabilities. It's not magic—it's an incredibly sophisticated but understandable process of pattern recognition, feedback incorporation, and iterative improvement.
This demystified AI for me in a way that would prove crucial for my later work. Instead of seeing AI as a black box, I understood it as a powerful tool that could be guided, improved, and effectively collaborated with.
Building Confidence with AI Systems
Perhaps most importantly, this role gave me confidence working with AI that many developers still lack. While others approach AI tools with uncertainty or skepticism, I understood them as powerful collaborators that could dramatically enhance my capabilities when used effectively.
This confidence would prove crucial when I later built 65+ applications using AI-assisted development techniques.
Reflections on a Transformative Experience
The Value of Stepping Outside Your Comfort Zone
Taking on AI mathematics training wasn't an obvious career move for someone focused on software development. But it ended up being one of the most valuable experiences of my professional journey.
It taught me that sometimes the most important learning happens when you step outside your expected path and dive into something completely different.
Understanding AI from the Inside Out
This experience gave me a perspective on AI that most developers don't have. Instead of just using AI tools, I understood how they were built, trained, and refined. This inside view made me a more effective AI collaborator and a more thoughtful evaluator of AI capabilities.
The Importance of Human-AI Collaboration
Most importantly, this role taught me that the future isn't about AI replacing humans—it's about humans and AI working together more effectively. The best results come from combining human creativity, judgment, and contextual understanding with AI's computational power and pattern recognition capabilities.
Looking Forward: The AI-Enhanced Future
My time training AI models at Outlier was brief but transformative. It gave me skills, perspectives, and confidence that would prove essential for everything that followed. It also gave me a deep appreciation for the potential of human-AI collaboration.
The mathematical problems I helped AI systems learn to solve were just the beginning. The real opportunity was in applying those same collaborative principles to software development, problem-solving, and innovation.
The future belongs to people who can work effectively with AI. And sometimes, the best way to learn that is by teaching the AI first.
This post is part of my "Professional Journey" series, where I share authentic experiences from my career path. Next up: diving deep into full-stack development at eCenter Research.