The Next Wave of AI Security Is Opening Doors for Data Professionals

AI is embedded across the curriculum, faculty research and in teaching practice in the UC Davis Master of Science in Business Analytics program. The San Francisco location and proximity to Silicon Valley is a big advantage as we are often treated to hearing first-hand from executives leading the AI and data science revolution.

Recently, Google DeepMind Senior Research Engineer Vaibhav Tulsyan joined our UC Davis MSBA Machine Learning and Artificial Intelligence course and, for a few hours, the future of AI felt a lot less abstract.

Tulsyan has published multiple papers in the AI space and has a close connection with our Assistant Professor Jörn Boehnke. Tulsyan shared about his career path, the evolution of large language models and the work currently shaping the field. 

What stood out immediately was how directly his experience connected to what we are learning in the MSBA program, not as theory, but as something actively being built and refined in real time.

Tulsyan began his work with Google in 2019 as a senior software engineer, where he was a key element to the foundational large language models that Google still uses today. Four years later, he transitioned to Google DeepMind, focusing primarily on post training of Gemini’s coding abilities as well as building Big Sleep, an AI security agent that was designed to locate critical bugs within open-source projects.

Image

After covering his career, Tulsyan gave us a crash course in large language models, “LLMs 101”. There, he explained to us the four main steps of LLM creation: 

1. Pretraining
2. Supervised Fine Tuning
3. Reinforcement Learning
4. Reinforcement Learning with Verifiable Rewards (RLVR)

Tulsyan briefly explained each phase and stressed the importance of RLVR as models become advanced. 

I’m currently working with data in a more applied setting, so this part of his talk stood out to me. In my role as a Graduate Student Analyst with the MSBA careers team in San Francisco, I spend much of my time collecting and analyzing alumni job data and opportunity trends, often using tools like web scraping to build and structure datasets. Hearing how similar concepts scale in real-world AI systems added another layer of context to the work I’m doing.

The session then moved into a Q&A, where Tulsyan shared perspectives on what it means to build a career in today’s AI-driven landscape.

• Software generation is disappearing, but that’s not a bad thing.
  • He emphasized that although routine code work within languages like Python or Java could be done by LLMs, this then freed up engineers and data to focus on novel areas of technology and expanding the areas around them.
• An AGI (artificial general intelligence) world still needs to be maintained by engineers.
  • As AI continues to evolve, so does the need to ensure it behaves as intended. He talked about the core pillars of AI safety—alignment, interpretability and control—stressing that these pillars are fields which will grow in size as AI expands its capabilities.
• Entry-level roles may be hard to land now, but they’re not disappearing.
  • As an engineer, researcher or scientist gains experience, they gain bias. Tulsyan states simply that all companies value the unbiased opinions of fresh graduates. Additionally, with the assistance of LLMs, some companies are finding that fresh graduates can solve some problems just as well (if not better) than those who are five to 10 years their senior.

Hearing his thoughts as a prospective data scientist was incredibly motivating. The reassurance and insight he provided into the future of not only software engineers but the tech world in general left me with hope for the future.

Experiences like this are what stand out to me about the MSBA program. The ability to hear directly from someone working at the forefront of AI, and to connect that back to both coursework and applied work, creates a clearer link between what we are learning and where it can lead.

For me, that connection is becoming more tangible, not just in the classroom, but in the work I’m doing alongside it.