Compensation: $200+ base (DOE) + performance-based incentive + options + excellent healthcare and life insurance, 401k+ benefits, PTO
Monarch AI is hiring a Principal-level AI Engineer to join our rapidly growing agentic AI vertical SaaS company, where we are uniquely positioned to dominate a number of vertical sectors serving building products & materials manufacturers. We're looking for Engineers who are exceptional builders of AI agents, who have built a 0-1 multi-tenant SaaS product before, and who ideally has enterprise, experience shipping production-grade AI agents at scale. A data science background is also a major plus.
This role is for someone with proven experience building enterprise-grade, production-ready AI Agents and multi-agent LLM systems, with deep expertise in how AI agents behave in enterprise environments — and who has personally built and deployed multiple AI agents and/or LLM-based systems into production.
We're looking for an individual contributor who wakes up and wants to build innovative AI Agents and valuable Agentic software each and every day. This is a work-hard/high-reward position where exceptional work is rewarded.
We are looking for an engineer who has deep experience in:
- Building enterprise production-grade AI agents
- Has shipped multi-tenant software as a major individual contributor
- system prompt architecture, and multi-step agent design
- tool-calling reliability, retries, fallback logic, and failure handling under real-world conditions
- eval frameworks, regression testing, benchmarking, and how to measure agent quality over time
- agent memory architecture, including short-term vs. long-term memory, retrieval-based memory, state persistence, memory relevance filtering, and the tradeoffs between useful memory, noisy memory, and dangerous memory corruption
- agent state management, session continuity, workflow state, and how to preserve context across complex multi-step tasks without creating brittle behavior
- hallucination prevention, edge-case handling, and designing systems that remain trustworthy when models are uncertain, wrong, or inconsistent
- structured outputs, schema enforcement, validation layers, and techniques to ensure model responses can safely drive downstream actions
- retrieval design, context management, chunking strategy, ranking, context injection, and how to maximize useful recall without overwhelming the model
- model quality, latency, and cost tradeoffs, including when to use stronger models, weaker models, routing layers, and staged reasoning patterns
- observability, tracing, logging, and how to inspect agent decisions, tool calls, and failure points in production
- guardrails, approval flows, permissioning, and human-in-the-loop controls for high-risk or irreversible actions
- orchestration of multi-agent systems, including delegation, task decomposition, coordination logic, and avoiding unnecessary agent complexity
- context window limitations, token budgeting, and designing around model memory constraints in practical production environments
- reliability engineering for LLM systems, including idempotency, retry safety, partial failure recovery, and graceful degradation
- model drift, prompt drift, and how to maintain system performance as models, workflows, and user behavior evolve over time
- real-world production tradeoffs between agent autonomy and deterministic software control
We want the kind of AI engineer who knows how to turn LLM capability into systems that are durable, testable, production-grade, and truly enterprise-ready.
- Ideal background: Proven experience building and deploying multiple production-grade AI agents and/or LLM systems that have operated in real-world environments, not just prototypes or internal demos
- Strong hands-on experience with enterprise-ready AI systems, including reliability, scalability, maintainability, observability, and safe deployment practices
- able to think deeply about agent architecture, workflow design, memory, orchestration, tool usage, and state management — not just prompt writing
- understands where and why agents break in production, including hallucinations, tool failures, context loss, memory corruption, brittle workflows, and edge-case behavior — and knows how to engineer around those realities
- experienced building durable AI systems with proper evals, regression testing, guardrails, structured outputs, validation layers, logging, tracing, and iteration loops
- understands how to balance model quality, latency, and cost while still delivering strong real-world performance at scale
- able to design systems that are not only intelligent, but also testable, debuggable, auditable, and trustworthy in an enterprise setting
- comfortable building from zero in ambiguous environments, while still bringing structure, engineering rigor, and clear technical judgment
- has high standards, low ego, and fast execution speed — someone who cares about getting the system right without becoming overly academic or slow