EPulseAI – Scaling Governed AI Across Engineering Teams

EPulseAI was a healthcare-focused generative AI platform created to help product and engineering teams incorporate AI into their day-to-day workflows. My work focused on designing a governance and prompt management experience that enabled teams to discover, review, reuse, and share effective prompts across the organization. It was demonstrated at Emids Healthcare Summit 2023 as a production-ready approach to enterprise AI adoption.

Software developers collaborating at laptops while building AI-assisted product development solutions in a modern workspace.

The context

When tech teams first started adopting generative AI, engineers naturally wanted to use it to speed up how they built software. However, because there were no team-wide guidelines, the rollout quickly became messy:

High Operational Risks: In a strictly regulated industry like healthcare, this unorganized approach wasn’t just inefficient—it was highly dangerous. It opened up severe security holes, risked breaking strict industry privacy laws, and threatened corporate compliance

Scattered Knowledge: Every engineer was writing their own AI prompts. These valuable instructions were hidden away in private text files and personal documents instead of being shared.

Unpredictable Results: Because everyone used different prompts, the quality of the AI’s output varied wildly from person to person. There was no reliable way to reuse the formulas that actually worked.

EPulseAI started as an idea: help enterprise engineering teams use generative AI more consistently.

The challenge wasn’t simply designing an interface for prompting. We had to figure out what an enterprise prompting system needed to support—how prompts would be created, reviewed, reused, governed and brought into developers’ existing workflows.

I helped turn that concept into a working product spanning a web platform and VS Code extension.

THE CHALLENGE:

Engineers were using generative AI in unorganized ways, creating major security and legal risks in a highly regulated industry.

MY ROLE:

Designed the entire experience for two connected tools: a central web dashboard for management and a coding plugin for developers.

THE OUTCOME:

Successfully launched the software and featured it at a leading healthcare technology summit as a real-world model for secure AI use.

Scope, ownership & design approach

My Scope

Led the end-to-end product design across two connected surfaces: a Web Application for managers to oversee prompt rules, and a VS Code Plugin for engineers to use those prompts while coding.

My Responsibilities

Owned the interaction architecture and visual design, mapped out secure workflows for regulated medical data, and simplified complex AI features for everyday users.

The core design principle

Prompting can’t just be a one-time chat. For tech teams, it needs to be a safe, repeatable workflow with built-in rules and clear tracking.

Key Design Decisions

Organize prompts by readiness

We grouped prompts by their current stage so teams instantly knew which ones to trust:

  • Used – Approved prompts ready to use in the coding tool.
  • New submissions – Developer prompts waiting for manager review.
  • Flagged – Prompts that need fixes before anyone can use them.
Track usage and adoption

Instead of guessing product quality, we provided a dashboard that tracks real developer habits:

  • Total prompt numbers, usage rates, and new submissions.
  • Additional Average effort needed for prompt refinement
  • Searches with no results, revealing unmet needs
  • Top contributors by team and individual
Create role based review access

We created explicit permissions to decide exactly who can submit, review, and approve prompts. This kept the process secure without creating a bottleneck for the engineering teams.

Key trade-off: User Freedom vs. Security Control

To keep data safe, the easiest design choice would have been to block certain words or restrict what developers could type into the AI. However, forcing strict input rules would slow engineers down and cause them to stop using the tool entirely.

Instead, I chose to govern the lifecycle of the prompts rather than restrict the input itself. Developers kept their complete creative flexibility while writing code, while the company still achieved total security oversight by reviewing and approving the final prompts before they were shared team-wide.

Roles and permissions identified for the platform
Site map illustrating how the EPulseᴬᴵ web app structured information to support prompt governance—covering metrics, prompt lifecycle management, and team access for VS Code usage.

Interaction & Visual Design

Prompt lifecycle showing how free-form prompts move from VS Code submission to review and approval in the EPulseᴬᴵ web app, before being made available for team-wide use.

I started by designing the main governance dashboard to establish a clean visual baseline for the platform. Working closely with engineering and product partners, I focused on displaying only the critical indicators needed to evaluate adoption, tracking metrics, and security risks.

Through rapid wireframing and interactive reviews, I kept the interface highly focused—ensuring it gives leadership actionable data insights without overwhelming them with unnecessary screen clutter.

Dashboard design surfacing prompt adoption and usage across teams. A card-based layout was chosen to support scalability and future customization, with ordering managed via backend configuration in the MVP.
Chat interface designed for product and delivery managers, combining conversational interaction with persistent visibility into prompt search results and recent activity to support informed reuse and continuity.
Selecting a search result auto-populates the prompt input, with a subtle visual indicator linking the selected result to the populated text to reduce confusion and maintain context.
Prompt management interface designed to support governed reuse at scale—enabling teams to filter prompts by status, switch contexts across teams, and review detailed prompt information without leaving the workflow.

Designing for VS Code plug-in

Alongside the web app, I designed a native-feeling side panel to bring approved AI prompts directly into the developer’s daily workspace.

The challenge

The VS Code developer framework has strict layout limits and intentionally restricts visual customization to preserve coding speed.

The design strategy

Instead of fighting these technical limits, I treated them as helpful boundaries. I aligned every interaction, button, and menu with native VS Code design guidelines.

The impact

Prioritizing familiarity and speed allowed developers to search, explore, and run prompts seamlessly without breaking their coding workflow.

Basic search functionality. If user wants to search using more attributes, they can click on advance search.
VS Code plugin UI designed as a lightweight side-panel experience, enabling prompt search and exploration while adhering to native VS Code patterns for speed, familiarity, and minimal disruption.

Impact – From concept to a working product🚀

EPulseAI evolved from an idea for improving AI adoption into a working product spanning a governed prompt management platform and a VS Code extension. The design established a structured way for engineering teams to create, review, manage, discover and reuse prompts—while giving teams visibility into how prompts were being used.

The product was demonstrated at the Emids Healthcare Summit 2023, showing how AI experimentation could move from individual use toward a more structured enterprise workflow.