Meet Leon Qi

Meet Leon Qi, Founding AI Engineer of actAVA.ai

Leon Qi is an AI systems engineer specializing in agentic architectures, retrieval-augmented generation (RAG) pipelines, and production-grade reasoning services. At Salesforce, he served as a Senior Software Engineer across Industries Cloud, AI Research, and Agentforce, where he helped architect LLM-powered reasoning engines, orchestrated multi-agent workflows, and delivered enterprise-grade AI agent systems with robust reliability, security, and observability practices. His work has focused on building intelligent systems that combine advanced language models with structured reasoning capabilities to solve complex enterprise challenges.

Before this, he was a Senior Software Engineer at Vlocity (acquired by Salesforce in 2020), where he contributed to the Vlocity Mobile Platform by developing hybrid mobile frameworks and reusable UI components for iOS and Android field-service applications. His solutions enabled frontline workers to access critical business data and execute complex workflows on mobile devices. He brings comprehensive full-stack expertise spanning web, mobile, cloud infrastructure, and AI/ML systems, with early-career experience delivering large-scale enterprise software and modern web applications on a variety of platforms.

When you say AI engineering is both harder and easier than traditional programming, what exactly do you mean?

It's a great paradox, and I think about it a lot. On the "easier" side, AI has fundamentally changed what a small team can accomplish. With large language models, you can build something in a weekend that would have taken an entire engineering team months to prototype just a few years ago. You describe what you want in natural language, the model reasons about it, and you get a working output. The barrier to entry for building something with AI has never been lower.

But here's the catch, and this is the "harder" part. Getting from a working demo to a production system that enterprises can actually trust is an enormous leap. In traditional programming, if you write a function, it behaves the same way every time. With AI, the same input can produce a different output depending on context, model state, and a dozen other variables. Now add healthcare into the mix, where a wrong answer isn't just a bug; it could affect patient care. You start to understand the complexity. You need layers of guardrails, observability, compliance, and testing that simply don't exist in traditional software engineering. The code itself might be shorter, but the architecture around it, the reasoning traces, the governance, the safety nets, that's where the real engineering challenge lives.

Why does it make you believe actAVA's approach to AI engineering has enabled it to create its agentic platform?

What I saw at large enterprises is that most organizations try to bolt AI onto existing systems as an afterthought. They take a generic chatbot framework, connect it to a model, and hope for the best. That approach breaks down the moment you need real accountability, when you need to know why an agent made a specific recommendation, or when you need to guarantee that patient data never leaks between conversations.

actAVA took the opposite approach. We built KORA, our AI agent platform, from the ground up with healthcare and life sciences as the primary design constraint, not a secondary consideration. Every architectural decision flows from that. Our platform is organized into three integrated suites that cover the entire agent lifecycle: BLUE handles agent building and orchestration, so teams can create sophisticated AI agents through a natural language interface without needing deep technical expertise. RED is our testing and compliance suite. It continuously evaluates agents for accuracy, safety, and regulatory adherence, catching hallucinations, bias, and compliance violations before they ever reach a patient. And GREEN is our continual learning engine, which lets agents actually improve over time through a multi-timescale learning architecture rather than staying frozen after deployment.

What makes this work is that these aren't separate products stitched together. They're layers of a single platform where security, compliance, and intelligence are woven into every interaction. Each agent runs in its own isolated container. Think of it like giving every AI worker their own secure office where they can't accidentally access anyone else's files. That level of architectural intentionality is what lets us move fast without cutting corners on safety.

What has been your favorite feature to work on in the actAVA KORA product suite?

Honestly, it's the agent building experience inside BLUE. Specifically, how easy we've made it to go from an idea to a running agent. That was a deliberate engineering obsession for us.

In most platforms, building an AI agent still feels like programming. You're writing code, wiring up integrations, configuring infrastructure. We flipped that. With KORA|BLUE, you define your goal, lay out the tasks or plan, and select the tools to achieve it, all through a conversational, no-code interface. We handle the rest. A quality director who has never written a line of code can build a sophisticated agent that analyzes clinical data, pulls from multiple sources, and generates compliance reports. That's not a simplified chatbot. That's a deep, reasoning agent built by the person who actually understands the problem.

What makes me especially proud is what these agents can actually do once they're running. They're not limited to answering a quick question and disappearing. Our agents handle long-running, complex workflows. They can connect to dozens of tools, orchestrate multi-step processes, pull data from your systems, run calculations, generate documents, and keep working through an entire end-to-end task without losing context. Think of a single agent that can take a quality improvement initiative from data gathering through root cause analysis to drafting the action plan, all in one continuous session.

And the whole time, every agent runs in its own isolated container, a completely sandboxed environment with its own secure runtime. It can't reach beyond its boundaries, can't access another tenant's data, and can't cause unintended damage to the broader system. It's like giving each agent its own secure workspace with everything it needs to do its job, but nothing it doesn't. For healthcare organizations, that's the difference between experimenting with AI and actually trusting it in production. We made it easy to build, powerful enough to handle real work, and safe enough to sleep at night.

Bonus Round: What is your prediction for the biggest AI trend impacting your customers in 2026?

I think 2026 will be the year when "agentic AI" stops being a buzzword and becomes operational infrastructure. Right now, most healthcare organizations are still experimenting, running pilots, testing chatbots, dipping their toes in. But the organizations that will pull ahead are those that move from single-purpose AI tools to coordinated multi-agent systems capable of handling end-to-end workflows.

Imagine a world where an AI agent doesn't just answer a question about a quality metric. It identifies the trend, pulls the relevant clinical data, drafts the improvement plan, routes it for approval, and monitors the outcome over time. That's not science fiction; that's what platforms like KORA are already enabling. The shift from "AI as a tool" to "AI as a teammate" will accelerate dramatically, and the organizations that have the governance and infrastructure in place to trust their agents will have a massive advantage.

The other trend I'm watching closely is continual learning, the idea that your AI systems get smarter with every interaction rather than staying frozen in time. That's exactly what we built GREEN to do, and I think it's going to become a baseline expectation. Static AI will start to feel as outdated as static websites.

Thanks for your time today, Leon, and for helping us understand why it is so hard to make it look so easy!

Visit us at https://actava.ai/ to learn more.

Meet Leon: View LinkedIn Profile