On October 15, 2026, Microsoft and Carnegie Mellon University (CMU) announced their answer to that existential riddle. In an essay on Microsoft's Command Line blog co-authored by Microsoft Azure CTO, Deputy CISO, and Technical Fellow Mark Russinovich alongside Microsoft Vice President Scott Hanselman, the company unveiled PRAISE—short for Preceptorship for AI-accelerated Software Engineering. Rather than pulling up the career ladder behind today's veteran architects, Microsoft and CMU are testing a structured training model that pairs early-career Microsoft engineers with seasoned mentors and formal graduate-level coursework.
The Junior Developer Paradox in the Age of AI
To understand why Microsoft and Carnegie Mellon are intervening directly in early-career training, look at how software engineering has historically worked. For decades, junior programmers built their technical intuition from the ground up. As Michael Hilton, Teaching Professor and Associate Department Head for Education in the Software and Societal Systems Department at CMU's Kenneth C. Griffin School of Computer Science, noted in both the Microsoft and CMU announcements, early-career software engineers traditionally did a high volume of routine tasks that gradually built up their expertise.
Now, generative AI coding tools handle much of that low-level groundwork almost instantly. Industry observers have pointed out a looming side effect of this automation: as writers on Vocal Media's Futurism channel and RTInsights have noted, when AI tools continuously generate code suggestions, early-career developers risk skipping the trial-and-error struggles that build deep structural problem-solving skills, even as the job shifts toward reviewing whether code is secure, scalable, and maintainable.
According to Carnegie Mellon University's School of Computer Science, the spark for PRAISE came in October 2025 when Russinovich—himself a CMU alumnus—visited the Pittsburgh campus. During that visit, Russinovich observed that while AI tools provide substantial leverage to senior engineers, junior engineers do not automatically get the same benefit. A veteran architect already knows how to spot a subtle concurrency bug or a security flaw in machine-generated output; a newcomer might not know what questions to ask.
Section Summary:
- AI coding tools automate many of the routine tasks that historically helped junior developers build foundational skills.
- Mark Russinovich noted during an October 2025 visit to CMU that AI tools assist senior engineers far more naturally than they assist junior engineers.
- Without a way to train junior engineers today, the industry risks a shortage of capable senior engineers tomorrow.
Inside PRAISE: Borrowing the Nursing Preceptorship Model
Instead of leaving new hires to sink or swim with a prompt window, Microsoft borrowed a time-tested training structure from healthcare: the nursing preceptorship. In hospitals, newly graduated nurses are paired directly with experienced clinical preceptors to build real-world judgment, confidence, and accountability under supervision before flying solo.
In Microsoft's adaptation, early-career engineers—referred to as "preceptees"—are paired with senior engineering "preceptors." As Russinovich and Hanselman explain on Microsoft's Command Line blog, the objective is not to force newcomers to write code by hand as if AI had never been invented. Instead, the program aims to teach them how to use AI tools thoughtfully while cultivating systems thinking, engineering taste, and technical judgment.
How the Workplace Mentorship Functions
Within Microsoft's engineering teams, the preceptor-preceptee relationship focuses on active, deliberate learning rather than passive observation:
- Workshopping Problems in Safe Spaces: Preceptors set aside dedicated time for junior engineers to ask questions and work through architectural problems without the pressure of matching the speed and scale of multi-year veterans immediately.
- Interrogating Assumptions: Junior engineers are coached to test their assumptions and discuss unfamiliar code changes with colleagues rather than shipping AI-generated code sight unseen.
- Transferring Engineering "Taste": Preceptors walk through their own problem-solving approaches, explaining why one design pattern is preferable to another and helping preceptees decide where to focus their learning based on their current skill set.
Russinovich and Hanselman compare the rise of AI in software engineering to the historical shift toward compilers. Compilers abstracted away raw assembly language, making development faster without eliminating the need for engineers who understand how systems work under the hood. When the metaphorical car breaks down in production, someone still has to pop the hood and diagnose the engine.
Section Summary:
- PRAISE pairs early-career "preceptees" with senior "preceptors" inside Microsoft, modeling the relationship on clinical nursing training.
- Mentors focus on systems thinking, testing assumptions, and reviewing unfamiliar changes rather than letting AI output pass unexamined.
- Microsoft compares AI coding tools to compilers—an abstraction layer that changes daily workflow without removing human accountability for the final product.
The Carnegie Mellon Curriculum: What Preceptees Actually Study
Workplace mentoring is only half of the PRAISE equation. Rather than expecting either universities or tech companies to carry the entire burden of modernizing developer education alone, Microsoft partnered with CMU's Kenneth C. Griffin School of Computer Science to backstop on-the-job mentoring with formal academic instruction.
According to Carnegie Mellon University's announcement by Marylee Williams, selected early-career Microsoft employees enroll in online CMU courses and small-group discussions while working full-time. Take the example of Jace Watson, a full-time Microsoft software engineer based in Seattle who was profiled by CMU. Watson noted that while consumer AI tools were emerging during his undergraduate years, using AI was not a major part of his university curriculum. Once he arrived at Microsoft, however, AI tools were embedded in his daily workflow, prompting him to enroll in PRAISE to learn how to use them rigorously.
CMU's Software and Societal Systems Department designed a specific sequence of graduate-level courses for the PRAISE certificate program, directed by professor Jonathan Aldrich:
| Program Phase | CMU Course Title | Primary Educational Focus |
|---|---|---|
| Fall Semester (Current) | Machine Learning in Production | Incorporating machine learning models safely and reliably into production software systems. |
| Fall Semester (Current) | Preceptorship in AI-Augmented Software Engineering I | First half of the core sequence on applying AI tools responsibly during software development and bridging theory with daily Microsoft work. |
| Spring Semester (Upcoming) | DevOps: Engineering for Secure Development and Deployment | Building, testing, and deploying secure, production-ready solutions at enterprise scale. |
| Spring Semester (Upcoming) | Preceptorship in AI-Augmented Software Engineering II | Second half of the applied preceptorship curriculum focused on advanced engineering judgment and delivery. |
According to CMU, credits earned in these four courses can be applied toward a full Master's Degree in Software Engineering at the School of Computer Science. Martial Hebert, Dean of CMU's School of Computer Science, described the partnership as a way to bring academic training to working professionals outside the traditional classroom setting, while Aldrich suggested the program could eventually serve as a template for other companies—and CMU noted it hopes to expand the certificate program to additional students and potentially other industry partners in the future.
Section Summary:
- PRAISE participants take online graduate courses through CMU while working full-time at Microsoft.
- The two-semester curriculum covers Machine Learning in Production, a two-part Preceptorship in AI-Augmented Software Engineering sequence, and DevOps: Engineering for Secure Development and Deployment.
- Completed coursework can be credited toward a CMU master's degree in software engineering.
Critical Analysis: A Blueprint or an Elite Experiment?
Does a computer science degree still matter when anyone can type a natural-language prompt and spin up a basic application? Russinovich and Hanselman argue emphatically that it does, asking rhetorically whether anyone would seriously ship AI-generated code in a finished product sight unseen. As engineering time shifts away from typing every line of syntax and toward architecture, validation, and system design, understanding what the underlying code actually does becomes the primary safeguard against catastrophic outages and security holes.
Still, let's put on our veteran IT analyst glasses and look at what is—and isn't—proven here. PRAISE is a thoughtful, well-structured response to a real workforce dilemma, but right now it is an internal program at one of the richest technology companies on Earth, partnered with one of the most selective computer science universities in the world.
Neither Microsoft's Command Line post nor CMU's announcement discloses how many junior engineers are enrolled in the current cohort, what the selection criteria or funding terms look like, or whether Microsoft has collected comparative data showing that PRAISE preceptees catch more bugs or reach senior engineering milestones faster than peers outside the program. For a mid-sized enterprise IT shop or a regional software vendor that cannot afford to sponsor Carnegie Mellon graduate tuition or dedicate senior architect hours to structured clinical preceptorships, the question remains: how do they train their junior developers?
Even without a CMU partnership, however, engineering managers and Windows enterprise dev teams can adopt the core mechanics of the PRAISE model on Monday morning:
- Ban "Sight-Unseen" Pull Requests: Require junior developers to explain line-by-line what an AI-generated code block does and what assumptions they tested before merging it.
- Pair Review with Architecture Walkthroughs: Have senior engineers spend less time nitpicking syntax—which linters and AI already handle—and more time explaining why a service boundary, memory allocation, or security control is designed a certain way.
- Treat AI Tooling as Ongoing Education: Pair daily coding tools with structured learning on secure DevOps deployment and production machine learning, ensuring junior staff learn how code behaves after it leaves their local IDE.
AI is undeniably rewriting how software gets built, turning developers from syntax typists into system architects and code auditors. By launching PRAISE, Microsoft and Carnegie Mellon are making a pragmatic bet: the fastest way to keep AI-accelerated software from going off the rails is to invest heavily in the human engineers holding the steering wheel.