Kheops
Description
Kheops is an ambitious, proprietary AI platform developed by Tech Fly, built to host and manage custom Large Language Models (LLMs) and intelligent, autonomous AI Agents. While the user-facing and management interfaces are built with Flutter, the heavy lifting—the model serving, custom training pipelines, and agent orchestration—is handled by a specialized Laravel backend augmented with high-performance computing services (like GPU instances). The platform allows users to deploy domain-specific LLMs (fine-tuned on proprietary data) and orchestrate various agents (e.g., researcher agents, coding agents) that can plan, use tools (APIs, databases), and complete complex, multi-step tasks. Kheops provides the necessary infrastructure for observability, security, and continuous learning, transforming AI from a passive model into an active, automated workforce.
Services
Realisation date
10 novembre 2025
Client name
Plan of Action
The initial and most critical phase involved establishing the Specialized Laravel Backend Infrastructure dedicated to managing model serving, including robust APIs for inference, and integrating with high-performance computing resources (GPU clusters) for training and fine-tuning the custom LLM. This was immediately followed by developing the core Agent Orchestration Engine in the backend, which gives the LLM the ability to access tools (like Laravel database connectors or external APIs), plan multi-step actions, and manage short and long-term memory for the agents. Concurrently, the Flutter Management Dashboard was built to provide administrators with clear interfaces for model deployment, agent monitoring, and observability (tracing agent decisions and usage metrics). The final steps focused on rigorous security auditing of the AI infrastructure and implementing continuous feedback loops and testing to ensure the LLM and agents maintain accuracy and prevent unwanted behaviors, leading up to a strategic deployment at the start of our work week.