Senior Software Engineer, AI Agents; Autonomous Systems

Senior Software Engineer, AI Agents (Autonomous Systems) — Gaia

Gaia is building the next generation of  experiences using AI. This role focuses on designing and shipping agentic, autonomous software systems that can plan, act, evaluate outcomes, and continuously improve—driving real product impact, not demos. You’ll build production-grade AI agents that meaningfully enhance customer experience, operational efficiency, content intelligence, personalization, and discovery across  .

This is a builder role: you’ll move fast, iterate relentlessly, and own outcomes end-to-end—from problem framing and system design to deployment, observability, and continuous optimization.

• Architect and implement agentic AI systems that autonomously execute multi-step workflows (planning, tool use, memory, evaluation, refinement).
• Build and own production services in Python that orchestrate LLM-based reasoning, retrieval, tool calling, and safe action execution.
• Design autonomy loops: task decomposition, reflection/self-critique, reward signals, evaluation harnesses, and guardrails.
• Develop robust RAG pipelines for Gaia’s content ecosystem (semantic search, chunking, embeddings, reranking, citations, freshness).
• Create frameworks for agent reliability: testing, simulation, regression suites, red-teaming, and continuous evaluation.
• Implement observability for LLM systems: tracing, cost/latency monitoring, failure taxonomy, quality metrics, and incident response.
• Partner with product, design, and content teams to translate Gaia’s mission and user needs into autonomous capabilities.
• Optimize for performance and cost: caching, batching, model routing, quantization (where relevant), and prompt/system improvements.
• Ship continuously: build, measure, learn—tight loops, pragmatic decisions, and visible progress.

• Expert-level Python and experience building production services (APIs, workers, pipelines, orchestration).
• Deep knowledge of LLMs and agentic systems, including strengths/limits, failure modes, and practical patterns for reliability.
• Proven track record of execution: you ship, you iterate, you improve outcomes based on real signals.
• Strong “builder + owner” mindset: you take ambiguous problems, create clarity, and deliver results.
• Entrepreneurial mindset: bias toward action, comfort with uncertainty, high accountability, and strong product instincts.
• Solid foundation in mathematics, statistics, and data reasoning (you can quantify uncertainty, validate improvements, and avoid hand-wavy conclusions).
• Strong data fluency: instrumentation, metrics design, experiment analysis, and operational decision-making using data.

• Hands-on experience building agentic workflows using modern frameworks (e.g., Lang Graph/Lang Chain, Llama Index, Semantic Kernel, or equivalent custom stacks).
• Experience with tool-using agents: function calling, structured outputs, constrained decoding, and robust schema validation.
• Experience with evaluation techniques for LLM systems (golden sets, model-graded evals, pairwise ranking, offline/online correlation).
• Experience with retrieval systems: vector databases, hybrid search, reranking, query rewriting, and content freshness strategies.
• Knowledge of prompt/system design for production (instruction hierarchies, routing, safety constraints, and jailbreak resistance).
• Experience with distributed systems and async execution patterns (queues, orchestration, retries, idempotency, back pressure).
• Experience deploying and scaling LLM-enabled services in cloud environments (AWS/GCP/Azure), including CI/CD and IaC.
• Familiarity with MLOps/LLMOps tooling: experiment tracking, model gateways, prompt/version management, and tracing.
• Experience with privacy/security considerations for AI systems (PII handling, data minimization, auditability).
• Front-end or full-stack capability is a plus (you can ship end-user impact, not just back-end components).
• Prior work in consumer subscription products, content platforms, personalization, or discovery systems.
• Designing clean architectures for autonomous agents: planners, executors, tool registries, memory stores, and evaluation…