About the RoleWe are building an agent that can build agents, "Factory for AI Agents." Our mission is to transform structured data and high-level goals into scalable, autonomous multi-agent systems that can execute complex tasks in real-world digital environments.We are looking for a visionary AI Engineer with deep expertise in architecting and orchestrating production-grade, multi-agent systems. The ideal candidate will be a master of agentic design patterns, from dynamic tool integration and adaptive memory to resilient coordination and human-in-the-loop safety.Core Responsibilities1. Core Agent Generation EngineDesign and build the core engine that translates structured inputs (APIs, data schemas, workflows) and prompts into functional, autonomous agents.Implement the logic to automatically analyze a given task and assign appropriate roles, reasoning strategies and memory scopes to newly generated agents.Utilize advanced metaprogramming techniques to write code that generates and configures agent dependency graphs and operational logic on the fly.2. Dynamic Instruction & Prompt ArchitectureEngineer sophisticated and modular instruction "blueprints" that serve as the DNA for agent behavior, defining their personas, operational boundaries, and required output formats.Design the "meta-prompt" logic that can interpret a user's high-level goal and compile it into a precise, machine-executable instruction set for a team of agents.Implement robust instructional guardrails and validation mechanisms to control agent behavior, prevent task deviation, and mitigate model hallucination.3. Multi-Agent Orchestration & ResilienceArchitect the runtime environment that governs agent-to-agent communication, enabling dynamic task delegation to specialized "sub-agents" to keep the primary agent's context clean and focused.Engineer resilient control flows with failure containment, including loop prevention, automatic retries, and checkpointing for long-running tasks.Implement sophisticated human-in-the-loop (HITL) workflows, allowing for human approval, editing, or rejection of sensitive or high-stakes agent actions before execution.4. Adaptive Memory & Learning SystemsDesign and implement a robust memory architecture that distinguishes between short-term (transient) task memory and long-term (persistent) knowledge.Build systems that allow agents to learn from user feedback and interaction, iteratively updating their own core instructions and knowledge base to improve performance over time.Structure agent memory to encompass procedural knowledge (how-to guides), semantic knowledge (facts), and episodic knowledge (interaction history).5. Pluggable Tool & Skill IntegrationDevelop a system for integrating external tools, from simple API calls to complex browser and desktop automation.Design a system for creating and managing modular, reusable "skills", on-demand capabilities that package specialized, multi-step workflows and can be dynamically loaded into an agent's context as needed.Create secure, sandboxed environments to ensure that all dynamically attached tools and skills are executed safely and without risk to the core system.6. Hybrid Model & High-Performance InferenceArchitect a flexible inference routing layer that dynamically selects the best model for a given task, balancing between powerful proprietary models (e.g., OpenAI, Anthropic, Gemini) for complex reasoning and fine-tuned open-source models for efficiency and speed.Deploy and manage high-throughput inference engines ( vLLM or similar) for open-source models.Optimize GPU infrastructure for maximum throughput, utilizing techniques like quantization, paged attention, and efficient KV cache management. Model layer and KV cache offloading between RAM, VRAM and disks for better memory management of open source LLMs .Technical Skills & RequirementsAgentic Design Mastery: Deep, hands-on experience building AI agents (not just chatbots) that take action. You have a strong command of agentic patterns like planning, reflection, tool use, and memory management.Multi-Agent Framework Experience: Proven experience with one or more multi-agent frameworks (e.g., LangChain, AutoGen, CrewAI, or custom-built solutions).Advanced Prompt Engineering: Expertise in designing scalable and dynamic instruction templates that govern agent behavior, define operational guardrails, and manage memory access.Memory & State Management: Experience designing and implementing agent memory systems, distinguishing between short-term context and long-term, persistent knowledge stores .Tooling & Automation: Experience building and integrating tools for agents, including API integration, schema transformation, and familiarity with browser or desktop automation frameworks High-Performance Inference: Experience with both commercial LLM APIs and deploying open-source models using high-performance engines. You understand the trade-offs between latency, throughput, and cost.System Architecture: Ability to design modular, extensible, and resilient agent architectures from the ground up.