The Challenge

AI-first games are a new frontier—games where generative AI creates content in real-time based on player actions. Our client had a compelling concept but needed a working prototype to demonstrate viability to investors.

The business model was novel:

• Players purchase the game cheaply, receiving initial LLM credits as virtual currency
• Gameplay consumes these credits (AI generates content in real-time)
• Players can purchase additional credits to keep playing

This meant we needed real commerce infrastructure: payments, virtual currency management, and mapping between real money and AI compute costs. All ready for a VC pitch in 2 months.

Our Approach

Build on Existing Code

Given the timeline, we built on the client’s existing Python APIs rather than rewriting. This minimized risk and let us focus on the critical new functionality.

Commerce Backend

We integrated Xsolla for payment processing and built the full monetization flow: real-world payments, virtual currency management, and LLM credit allocation.

AWS + Kubernetes Deployment

We deployed on AWS using Kubernetes for container orchestration, managed through Pulumi for infrastructure-as-code. AWS was already supporting the client with AI compute resources, making it the natural choice.

The Pivot

Two weeks before the pitch deadline, Xsolla deprecated their inventory product. We had to rapidly build a custom transactional inventory system using Redis to replace it.

The team delivered the replacement in two weeks, maintaining the timeline. When you have a VC pitch date, you don’t get to reschedule.

The Results

• MVP delivered within the 2-month window
• VC funding secured using the prototype
• Real commerce processing actual payments
• Virtual currency system mapping to LLM compute credits
• Rapid pivot when third-party dependency disappeared
• Ongoing relationship established for future work

This engagement demonstrated that when deadlines are immovable—like investor pitch dates—we find a way to deliver.