When most people see AI today, they see headlines splashed across billboards and social media feeds. But behind the scenes, AI is also quietly transforming the day-to-day work of professionals inside some of the world’s largest companies.
For Tianyu Zhang, that transformation has unfolded firsthand across three different stages of her career. From earlier workflows that relied heavily on manual analysis and operational coordination to today’s increasingly AI-augmented systems at Walmart, she has witnessed how artificial intelligence is reshaping not only technology itself, but also how people collaborate, make decisions, and solve problems at scale. As AI becomes embedded into everything from internal tools to enterprise operations, Zhang’s journey offers a grounded perspective on what this shift actually looks like inside the workplace.
As a software engineer, Zhang worked on large-scale enterprise applications supporting Walmart internal teams, merchants, and suppliers. She contributed to the development of GraphQL architectures, frontend and backend data integration, testing automation, and scalable engineering systems supporting millions of business operations. She was also among the early engineers exploring AI-assisted software development workflows within the organization, participating in experimentation around emerging technologies such as AI coding assistants and AI tools integrations.
Outside of her day-to-day engineering responsibilities, Zhang regularly shared technical knowledge with fellow engineers through internal technical presentations and broader engineering communities, focusing on practical strategies for effectively implementing AI tools into modern software development workflows.
Zhang holds a Bachelor of Science in Computer Science and Applied Mathematics from the University of Auckland and a Master of Science in Security Informatics from Johns Hopkins University. Her experience across software engineering, AI adoption, and cybersecurity provides a unique perspective on how AI is transforming the future of enterprise technology.
Before AI Era: Human-Centric Development
In the initial chapter of Tianyu’s tenure at Walmart, she served as a software engineer on the Shop With Friends team, where she focused her technical expertise on the engineering and development of the initiative’s foundational AI tools. Through the Shop With Friends (SWF) initiative, Walmart bridged the gap between digital convenience and personal connection, allowing shoppers to solicit fashion advice from their social circles without the need for a physical store visit or a formal membership. By selecting virtual models that mirror their own likeness, users could curate ensembles in a digital fitting room and seamlessly distribute links or QR codes across text and social platforms. During her first tenure, AI coding tools such as GenAI Playground (AIStudio) were primarily introduced as experimental tools, and had not yet matured enough to help engineers build at scale or generate production-ready code. Responses could be totally off-topic or hallucinating, but it was a good way to get feedback data for early-day model training.
Despite these limitations, Zhang quickly recognized the potential of AI to reduce repetitive engineering work and accelerate software development workflows once the technology matured. At the time, however, architecture design, implementation, debugging, and production readiness still relied almost entirely on human engineers.
First-Phase: Buggy and Unstable
In Tianyu’s second phase at Walmart, AI tools began evolving from unstable experimental assistants into practical engineering collaborators. As she transitioned into a new role focused on developing AI-powered inventory and shelf-recognition tools for store operations under the Modular Intelligence team at Data Ventures Org, she also noticed a major shift in how engineers interacted with AI in their daily workflows. Modular Intelligence (as referred below by image 2) allows business to leverage modular and shelf level images and insights to ensure our customer always finds the right product at the right location in store reducing customer friction.
Additionally insights provided by modular intelligence products will allow Walmart’s Merchant teams to work more closely with suppliers to identify gaps with existing mods that results in customer friction, nil picks and lost basket opportunity (image 3).
During this time, despite working on an AI-powered retail analytics B2B platform herself, she shared that AI coding tools were considered buggy and unstable at the time. “We had a b2b cursor subscription at the time, and required an employee ID/company email account, but I kept running out of the usage quota before the end of the month. It didn’t have the memory at the time so it makes fixing bugs quite challenging, I would still have to go through all the files and figure out the context for a proper bug fixing.” Tianyu shared, “One upside was that Walmart as a corporation offered internal learning opportunities and organized AI coding tool knowledge sharing sessions.”
Second-Phase: Growing Trust and Structural Code
Closer to the end of her second role, AI coding tools started to become part of engineers’ workflow. While AI could now generate cleaner and more structured code, Tianyu noted that contextual reasoning remained a challenge. In highly specific engineering scenarios, AI would occasionally “overcorrect” functioning logic and introduce solutions that appeared technically sound, but failed under the realities of the actual business context. An interesting example will be that AI sticks strictly to “1+1 = 2”, this kind of mathematical truth, regardless of the fact that the actual business context counters the mathematical truth. However, AI became increasingly valuable in areas such as code reviews, documentation, and especially unit testing.
Tianyu shared that AI-assisted test generation was particularly impressive at identifying edge cases engineers might otherwise overlook, significantly reducing review and debugging time. During this time, Walmart also started to invest and develop internal tools such as Code Puppy. Code Puppy appears to be an internal web-based or agentic platform, distinct from a standalone IDE, which focuses on “agentic AI” workflows. This tool is designed to generate code and articulate its reasoning, effectively enabling a more fluid and intuitive “vibe coding” experience for engineers.
Recent-Phase: Deep Integration Within The IDE
Most recently, Tianyu Zhang supports the Ship With Walmart team within Walmart’s Seller Center organization. Unlike her previous roles focused directly on consumer-facing and operational AI systems, her most recent work centers on building large-scale shipping infrastructure for Walmart’s e-commerce ecosystem.
The initiative enables seller-fulfilled marketplace merchants, as the vendors, that manage and store their own inventory to leverage Walmart’s transportation network as an alternative to traditional carriers such as UPS, FedEx, and USPS. By extending Walmart’s internal logistics capabilities from first-party operations to third-party marketplace sellers, the platform aims to improve delivery speed, strengthen on-time performance, and maintain cost competitiveness across the fulfillment ecosystem.
Although Tianyu has transitioned away from directly building AI products such as Shop With Friends and Modular Intelligence, she emphasized that AI has become even more embedded in her day-to-day engineering workflow.
She shared that Walmart has continued investing heavily in internal AI tooling, including a newer AI agent platform called Wibey. Similar to tools such as Cursor, Wibey integrates directly into the development environment and allows engineers to interact with large codebases through natural language prompts powered by AI models (e.g., Claude by default).
With a single prompt, developers can scaffold services, generate unit tests, resolve compliance issues, debug CI/CD pipelines, and automate repetitive engineering workflows without needing deep familiarity with every underlying repository or system dependency.
Unlike many external AI coding tools that impose strict usage caps, Tianyu explained that Wibey operates more like an internal engineering agent platform. When conversations approach token or context limitations, the system compresses prior context to free memory and continues the workflow rather than abruptly terminating the session.
According to Tianyu, tools like Wibey are rapidly reshaping the software engineering landscape by lowering the barrier to development and dramatically accelerating productivity. Engineers are increasingly able to execute tasks that previously required extensive onboarding or specialized system knowledge. Despite the progress of such, Zhang emphasized that engineers still need to validate generated code carefully, especially when business logic depends on company-specific context.
Tianyu’s View on AI Implementation In the Future for Software Engineers
When discussing the future of AI in engineering, Tianyu Zhang smiled and shared a perspective that felt both optimistic and reassuring for the next generation of engineers entering the industry. “I always think AI will increasingly become a core tool in software engineering rather than simply a replacement of engineers. With AI-assisted development, a simple engineer can become a super engineer who takes on multiple roles across development, testing, system operations, security, and project coordination.”
Rather than viewing AI as a threat to technical careers, Zhang believes the technology will elevate the role of engineers and shift the industry toward higher-level system thinking and orchestration: “As repetitive coding tasks become automated, future demand will shift more toward system architects, and engineers who can effectively leverage AI tools to design, validate and orchestrate complex systems.”
