Joe Lang

Our on-demand simulations exhibited a significant latency gap between the user's speech and the AI avatar's response, which degraded the sense of immersion and conversational flow.

I analyzed timestamped logs to identify a 1.2-second delay caused by Voice Activity Detection (VAD) settings; I initially optimized this to 0.5 seconds before migrating our Speech-to-Text (STT) stack to Deepgram Flux to leverage integrated VAD.

We achieved a total response time reduction of approximately 1.5 seconds, resulting in a consistently more responsive and immersive simulation experience for end-users.

Mursion’s simulation platform was built around live, human-led experiences. To expand the product into true 24/7, on-demand delivery, we needed a new technical architecture capable of running realistic simulations autonomously—while complementing the company’s existing human-led offering.

Working closely with an AI engineer, we architected and built Mursion’s first on-demand AI simulation platform, integrating real-time speech, generative AI, automated simulation logic, and responsive 3D avatars into a production system. I led the technical work required to connect these components into a low-latency experience that could operate without a live facilitator.

The platform created a new 24/7 delivery model alongside Mursion’s human-led simulations. AI simulations became the standard experience used in sales demonstrations and grew into a core product offering delivered to customers globally.

Mursion’s streaming architecture created a browser security boundary that prevented keyboard input from reaching the simulation, blocking keyboard-only navigation and creating a significant accessibility gap.

I partnered with our streaming provider to design and prototype a new API for securely forwarding keyboard input across the streaming boundary, then implemented the client-side integration to capture, transmit, and process those events within the simulation.

The solution enabled keyboard-only interaction across our streamed simulations, removing a key accessibility barrier and bringing the platform into alignment with keyboard accessibility requirements under WCAG 2.1.

As our AI simulations moved beyond the prototype stage, issues with conversation quality, animation, latency, and simulation logic became increasingly difficult to reproduce and diagnose. We lacked a systematic way to connect user-reported problems with the AI logs, video, and runtime data engineers needed to identify the root cause.

I designed and built an in-simulation Jira reporting pipeline that captured the context needed to investigate each issue and trace it back to the underlying system behavior. I automated the collection and attachment of simulation video, AI logs, and diagnostic data, creating an end-to-end workflow from issue report to engineering investigation. I owned and evolved the pipeline until it was mature enough to transition to QA for ongoing operation.

The system gave engineering and QA a repeatable way to capture, reproduce, diagnose, and resolve issues across our AI simulations. This feedback loop helped move the product from an early prototype toward a reliable, scalable offering while giving QA the tooling to independently manage AI simulation quality.

As Leia grew over the years, the Unity Plugin became increasingly more complicated to handle old requirements and new.

I simplified our Unity plugin logic from 40 classes to 8 without breaking core functionality. Additional features were then prioritized and mapped for integration.

We now had a plugin that was simple enough to reshape for future needs without the overhead of old assumptions and tech debt.

After Leia acquired Dimenco, we needed one plugin that could work across all displays on Windows and Android.

With our new simplified plugin structure, I was able to abstract out all device logic and set two streams for firmware communication.

We now have one logic flow that worked on all supported devices, Android or Windows. Later, when we added features, no code needed to be duplicated.

Leia's Unity plugin releases were ad-hoc, breaking changes were commonly thrown into release updates, and bugs were not caught prior to releases.

• Daily standups were introduced to review tickets and status
• Jira tickets were connected to branches and Pull Requests, making context and code changes easy to navigate
• Pull request workflow was standardized, encouraging team collaboration
• CI generated test builds to verify PRs did not break core functionality
• I mapped out our release process and announcement process so anyone on the team can launch an update

The workflow for the entire team was smoother, bugs were caught sooner, and the SDK was matured.

Our automotive partners needed to test rendering possibilities on experimental displays.

I built a process for pulling display configuration data with configurable JSON files. With that data, I could make builds for these displays without official firmware updates.

I was able to make demos for our automotive partner's experimental hardware and support them with internal projects.

Leia Unity SDK only supported default rendering, blocking content from working on our displays.

I restructured our SDK to add catches and hooks into Post Processing, URP, and HDRP rendering pipelines.

Leia Displays work out of the box with all rendering techniques.

Our plugins were facing several critical issues that were affecting customer satisfaction and hindering the adoption of our technology.

I led a team of three engineers to address and resolve these issues promptly. Our goals were to fix existing bugs reported by users, introduce new features requested by our clients, and manage the release of internal content related to the Unity Plugin to ensure it met our quality standards and user expectations.

Our team successfully resolved all identified bugs within the projected timelines, and we introduced several key features that significantly enhanced the functionality and user experience of the Unity plugin.

Industry partners and business stakeholders needed to showcase demos with various devices not available on the market, including experimental automotive hardware.

I was the go-to liaison who scoped the needs of the project and unblocked any barriers to creating content.

Leia was able to foster stronger relationships in various industries and continue support for experimental devices.

Leia did not have a standardized communication process for letting developers know about updates.

• Worked closely with the marketing team to develop clear, informative content about the new features
• Forum Engagement: dedicated a portion of my daily schedule to actively monitor and engage in our developer forums
• Feedback Loop: established a systematic approach to gather and categorize feedback from the forums, including bug reports and feature requests
• Update Turnover: set up regular meetings with our development team to discuss the community's feedback and prioritize actions

This process encouraged active community engagement in driving product success and customer satisfaction.

Leia embarked on an ambitious project to launch our marketstore with over 40 games. These games varied in complexity and audience, requiring a nuanced approach to porting, testing, and launching to ensure they met our high-quality standards.

• Project Management: developed a detailed project plan outlining key milestones and deadlines for each game, using project management software to track progress, assign tasks, and manage resources
• Communication: established clear lines of communication with our publishing partners via regular update meetings
• Quality Assurance Collaboration: worked closely with our QA team to develop a rigorous testing protocol for each game
• Feedback Loops: created a structured process for sending update requests and bug reports to our publishing partners
• Launch Coordination: coordinated with publishers to ensure a smooth launch as games passed QA

This comprehensive approach led to the successful porting, testing, and launching of all 40 games within the projected timeline.