Lead AI Architect

Postulation uniquement en ligne - merci de mentionner sous source (ORP)

Kandou is looking for an Lead AI Architect to help design, build, evaluate, and deploy advanced AI agent systems for real-world use cases. This role is focused on agentic systems that go beyond conversational assistants: complex analytical workflows, knowledge-based reasoning systems, controlled inference pipelines, tool-using agents, and transparent decision-support architectures.

Hands-on experience across multiple agentic AI projects, ideally spanning both industrial and academic environments. Should be comfortable working at the intersection of large language models, symbolic reasoning, knowledge representation, workflow orchestration, evaluation, and full-stack (web) product development.

This is a role for someone who can move from research concepts to working systems: designing agent architectures, implementing reasoning workflows, testing reliability, building user-facing interfaces, and ensuring that agentic behaviour is interpretable, controllable, and robust.

Key responsibilities:
Design and implement real-world agentic AI systems using modern agent frameworks and orchestration tools.
Develop agentic workflows that go beyond chat, including complex analytical pipelines, multi-step research workflows, tool-using agents, knowledge-grounded agents, and structured decision-support systems.
Work with knowledge-based AI architectures, including retrieval-augmented generation, knowledge graphs, symbolic rules, structured domain models, ontologies, and hybrid reasoning systems.
Develop and apply mechanisms for controlling inference, including planning constraints, reasoning policies, guardrails, validation layers, tool-use control, and human-in-the-loop checkpoints.
Explore and implement neuro-symbolic approaches for agentic reasoning, combining LLM-based reasoning with symbolic, rule-based, graph-based, or formally structured methods.
Build transparent AI methods that make agent behaviour traceable, explainable, testable, and auditable.
Create evaluation and testing frameworks for agentic systems, including benchmark tasks, regression tests, failure-mode analysis, trace inspection, robustness testing, and task-level performance measurement.
Develop full-stack prototypes and production applications, integrating backend services, APIs, databases, frontend interfaces, model providers, and orchestration layers.
Collaborate with researchers, engineers, product teams, and domain experts to translate ambiguous real-world problems into reliable agentic workflows.
Stay current with developments in agentic AI, reasoning systems, LLM orchestration, AI evaluation, and applied neuro-symbolic methods.

Required experience:
Strong multi-project experience developing real-world AI agents or agentic workflows.
Demonstrated focus on agentic reasoning, including planning, decomposition, tool use, multi-step inference, workflow execution, or autonomous task completion.
Experience in either industrial AI development, academic research, or ideally both.
Hands-on exposure to knowledge-based agentic systems, such as agents grounded in knowledge graphs, structured documents, domain rules, ontologies, databases, or retrieval systems.
Experience with methods for controlling reasoning or inference, such as guardrails, constrained planning, validation layers, policy-based tool use, symbolic checks, or deterministic workflow components.
Familiarity with neuro-symbolic AI concepts or hybrid reasoning architectures.
Experience designing transparent, inspectable, or explainable AI methods.
Practical experience with agentic reasoning evaluation, testing, benchmarking, observability, or failure analysis.
Full-stack web development experience, including backend APIs and frontend application development.

Technical Skills:
Strong Python engineering skills.
Experience with modern LLM and agentic AI frameworks, especially:
LangChain
LangGraph
OpenAI SDK / OpenAI Agents SDK
Retrieval-augmented generation systems
Tool/function calling
Multi-agent or multi-step workflow orchestration
Agent evaluation and tracing tools
Experience with backend development, APIs, databases, and cloud or deployment environments.
Experience with frontend technologies such as React, Next.js, TypeScript, or similar frameworks.
Familiarity with vector databases, graph databases, semantic search, structured data pipelines, or knowledge graph tooling.
Someone who thinks beyond prompt engineering. Should be experienced in the architecture of reasoning systems: how agents decide what to do, how inference is constrained, how knowledge is represented, how workflows are verified, and how complex AI systems can be made reliable enough for real-world use.

Qualifications and Portfolio:
Mature open-source contributions AND/OR.
Portfolio projects related to agentic AI, LLM systems, knowledge-based AI, neuro-symbolic reasoning.
Experience building AI systems in domains such as scientific analysis, enterprise knowledge management, decision support, research automation, legal/financial/technical analysis, or complex operational workflows.
Experience with production-grade AI system design, including observability, monitoring, testing, security, latency, cost control, and reliability.
Familiarity with human-in-the-loop systems, provenance tracking, workflow auditability, or regulated environments.
Experience integrating LLMs with external tools, APIs, databases, code execution environments, or analytical engines.
Desirable: publications (at main NLP/ML/AI conferences)

www.kandou.ai/careers/