AI Engineering

The podcast discusses Teresa's journey from a non-engineering background to becoming a full-time engineer, dedicating 60% of her time to engineering and 40% to writing. She began by developing her first AI product, an "interview coach," which evolved into a partnership with Vistily, a company specializing in Opportunity Solution Tree software. Together, they integrated Teresas AI tools, such as interview snapshot generation and solution tree creation, into Vistilys product suite, with plans to expand to an AI interviewer. Her work focuses on creating collaborative tools allowing teams to interact with AI-generated outputs, such as interview notes and solution trees, while leveraging Vistilys compliance frameworks to avoid managing infrastructure or security herself.

The discussion emphasizes technical challenges in AI product development, including mastering production-level engineering skills like automated testing, CI/CD, and optimizing AI-generated code for scalability. Teresa highlights the iterative process of improving AI tools through real-world testing, data analysis, and logging traces, contrasting this with superficial prompt adjustments. She also outlines her vision for AI-driven training, aiming to replace traditional courses with a personalized AI agent, "Teresa Bot," capable of coaching users in areas like interviews and business fundamentals. This tool uses retrieval-augmented generation (RAG) and embeddings databases, tested with real user feedback to refine performance.

Key themes include the importance of hands-on learning in AI engineering, using tools like Claude to bridge knowledge gaps, and the value of iterative experimentation over passive learning. Teresa reflects on overcoming insecurities through continuous learning and embracing curiosity to pursue unexpected career paths. She underscores the role of data science in product management, the normalization of not knowing everything upfront, and how AI democratizes complex skills, enabling individuals to build expertise without traditional prerequisites. The narrative also highlights the shift from traditional training models to dynamic, AI-powered coaching systems, emphasizing collaboration, scalability, and user-centric design.

• What if you built a niche AI product and partnered with a compliance-focused SaaS company to skip infrastructure overhead?

  • Concrete move: Develop a specialized AI tool (e.g., a "business fundamentals coach") and pitch it to a SaaS company with existing SOC 2/GDPR compliance frameworks.
  • Why now: The AI tooling market is fragmented, and SaaS companies are eager to integrate AI features without managing compliance. Your product can fill a gap in their ecosystem.
  • Expected upside: Leverage their compliance infrastructure to scale faster, while retaining IP ownership and revenue-sharing from licensing.

• What if you upskilled in AI engineering by prototyping a tool with RAG and embeddings, then iterated based on real user feedback?

  • Concrete move: Use Retrieval-Augmented Generation (RAG) to build a prototype AI coach (e.g., for interview prep) and test it with 50+ users in a Slack community.
  • Why now: Rapid prototyping with RAG is achievable with tools like Claude, and user feedback loops can refine your product before public launch.
  • Expected upside: Validate your AI products value proposition early, reduce development risk, and build a loyal user base for future monetization.

• What if you focused on a single AI coach role (e.g., "discovery coach") and optimized it for a specific workflow, then expanded?

  • Concrete move: Build a narrow AI coach for a high-impact task (e.g., generating opportunity solution trees) and integrate it with a collaborative tool like Vistilys platform.
  • Why now: Specialization allows you to dominate a niche use case, and partnerships with existing tools can accelerate adoption without building your own infrastructure.
  • Expected upside: Create a high-margin, scalable AI service that solves a specific pain point, then replicate the model to other coach roles later.

• Dedicate 60% of your time to engineering work and 40% to content creation to balance product development with knowledge sharing, as Teresa did while transitioning into a full-time engineering role.
• Leverage partnerships with established companies (e.g., Vistily) to access compliance frameworks (SOC 2, GDPR) and infrastructure, allowing you to focus on AI innovation without managing backend operations.
• Invest in learning production-level engineering skills (CI/CD, automated testing, scalability) through targeted courses or community resources, as Teresa emphasized the need to optimize AI tools for performance at scale.
• Prototype AI tools with real user data (e.g., test 100+ hypothetical user questions) and iterate based on feedback, as demonstrated by Teresas rapid prototyping of Teresa Bot using RAG and embeddings.
• Implement data analysis and logging in your AI products to identify performance gaps and improve outputs, mirroring Teresas focus on trace analysis and iterative refinement through real-world testing.