Position Summary:

• Kaseya is looking for an AI/ML Engineer to help build the intelligence layer of the Kaseya Intelligence Platform (KIP), an agentic AI platform designed to autonomously manage IT operations for Managed Service Providers (MSPs).
• This is a hands-on engineering role at the intersection of Generative AI, agentic AI, applied machine learning, and distributed systems. You will build production-grade AI agents that orchestrate vendor API calls, interpret PSA/RMM signals, make autonomous decisions, and operate reliably across complex MSP environments.
• You will also develop AI/agent evaluation frameworks, behavioral anomaly detection, safety mechanisms, and production workflows, while translating relevant ML/AI research into practical implementations. The role requires strong software engineering fundamentals and the ability to take AI concepts from experimentation through production.
• Build and iterate on agentic AI workloads that orchestrate vendor API calls, interpret PSA/RMM signals, and make autonomous decisions across MSP client environments
• Develop behavioral anomaly detection models that baseline per-workflow-identity call patterns, credential checkout volumes, and cross-vendor sequences: and flag deviations in real time
• Implement agentic-specific detection signals: tool call sequence anomalies, scope escalation attempts, decision branching inconsistencies, and output volume outliers
• Build harnesses and evaluation frameworks using tools like Amazon Bedrock, LangChain, and LangFuse to assess model quality, latency, and safety
• Read and translate research papers into production implementations: this role requires comfort navigating ML literature and turning concepts into working systems
• Contribute to connector codegen pipelines: agentic coding against vendor OpenAPI specs to accelerate Safe Connector Layer coverage across 40+ MSP tool APIs
• Work with the Temporal durable execution runtime to ensure agentic workloads behave predictably under failure, retry, and partial execution conditions

Required Qualification:

• 5+ years of experience in AI/ML engineering, software engineering, applied machine learning, or a related field, with strong hands-on development experience.
• Proven experience building and deploying production-grade LLM or agentic AI applications, beyond basic LLM/API experimentation.
• Strong programming and software engineering skills, preferably with Python, along with experience building scalable and reliable production systems.
• Hands-on experience with LLM/agent orchestration frameworks such as LangChain, LangGraph, LlamaIndex, or equivalent technologies.
• Practical experience with at least one major LLM/model platform, such as Amazon Bedrock, OpenAI, Google Vertex AI, or equivalent.
• Experience designing and implementing AI/LLM evaluation frameworks, including evaluation harnesses, benchmarks, test suites, quality measurement, latency analysis, or agent behavior evaluation.
• Strong understanding of agentic AI concepts, including tool/function calling, agent orchestration, multi-step workflows, decision-making, and autonomous execution.
• Ability to read and interpret ML/AI research papers and translate relevant techniques into working software or production implementations.
• Strong understanding of AI reliability and safety, including hallucination handling, guardrails, unexpected tool calls, behavioral anomalies, and evaluation of agent behavior.

Preferred Qualification:

• Experience building AI/ML systems for workflow automation, enterprise platforms, cybersecurity, security tooling, or IT operations.
• Experience with Temporal or other durable workflow/execution engines for reliable AI-agent execution.
• Experience with ClickHouse or other columnar/analytical databases for high-volume analytics and behavioral data.
• Experience building API/tool integrations, OpenAPI-based code generation, connector frameworks, or automated agent tool ecosystems.
• Experience with LLM observability, agent tracing, evaluation platforms, guardrails, and AI safety tooling, including tools such as LangFuse or equivalent.