Mar 2025 – Sep 2025 · 6 mos · Jaipur, Rajasthan, India · Hybrid

• When I joined Vivid AI in March 2025, the company was still in its ideation stage. We were a team of just three:- one frontend developer, one AI engineer, and myself responsible for cloud and infrastructure. My role was to research, design, and build a scalable architecture that could support the launch of our first products. After months of iteration and testing across platforms, we built a robust setup that enabled us to successfully launch Magic Swap and Virtual Try-On within six months.
• Key Contributions:
• Conducted research and benchmarking on GPU performance by comparing inference speed, resource efficiency, and cost-effectiveness of various GPU types (e.g., NVIDIA A100, RTX 4090).
• Prepared performance reports highlighting trade-offs between speed, stability, and pricing, which helped the team choose the best GPU setup for production deployment.
• Deployed AI diffusion models across multiple GPU platforms (AWS EC2 GPU instances and Lambda Labs) and tested how different environments perform in real-world workloads.
• Implemented a Redis-based job queue to manage user prompts and model responses, ensuring reliable request handling at scale.
• Integrated Slack triggers to notify the engineering team whenever a user visited the website or performed specific actions, improving team awareness and response time.
• Set up a CI/CD pipeline using GitHub Actions to automatically deploy code updates to AWS EC2 whenever developers pushed changes to the GitHub repository, reducing manual deployment effort and ensuring faster feature delivery.
• Implemented Prometheus and Grafana to monitor GPU utilization, system metrics, and performance trends.
• Key Skills: AWS (EC2, S3, IAM), VPC, Lambda Labs, Redis Queue, GPU Benchmarking, CI/CD with GitHub Actions, Slack Integration, Cost Optimization, Linux