AI Agent
AI DevOps & Infrastructure
Vikk AI: Real-time AI voice evaluation and POC
"Vikk AI partnered with Tech 42 to evaluate a flexible, cloud-native foundation for enhancing the real-time voice experience in our consumer platform. Tech 42 delivered a proof of concept that demonstrated how voice interactions, AI models, business tools, and knowledge retrieval could be organized into a modular and scalable architecture.
The engagement also included a structured evaluation of several AI and voice technology options, as well as performance testing to help assess the solution’s readiness for future production use.
The resulting architecture gives Vikk AI greater flexibility to evaluate and adopt different technology providers as the AI ecosystem evolves. It also establishes a clearer separation of platform components, which we expect will reduce future integration effort, accelerate development, and improve the resilience and maintainability of our voice capabilities.
The Tech42 team was professional, knowledgeable, and responsive throughout the engagement. They delivered what Vikk AI needed for this phase, and we look forward to continuing the partnership as we move toward production."
Haytham Allos
CTO
Project summary
Vikk AI provides conversational legal guidance to over 112,000 registered users through real-time voice and text chat. Their voice experience depended on third-party AI and voice providers, creating dependency and outage risk with limited flexibility to switch models. Tech 42 designed and built a proof-of-concept real-time voice AI agent on AWS. The architecture uses ECS Fargate with Pipecat Real-Time Media Server, a Bedrock Knowledge Base with OpenSearch for retrieval-augmented generation, Aurora PostgreSQL for conversation history, and ElastiCache for caching. Two full pipelines were evaluated across multiple speech-to-text, LLM, and text-to-speech options, with a structured framework measuring latency, cost, and conversational quality. The engagement delivered infrastructure as code, containerized services, and a knowledge transfer session, giving Vikk AI a documented foundation for moving toward production.
