Senior AI Engineer

Overview: Guidepoint is investing in applied AI across its research platform. We're building conversational AI systems that interact with our global expert network at scale — combining real-time voice, LLM-driven orchestration, and retrieval over large proprietary datasets. You'll be an early engineer on a new initiative: taking working prototypes to production and shaping the architecture as the product scope grows. This is a hands-on role with broad ownership, close to product and leadership. This role is hybrid in our Toronto Office What You'll Do: Real-time conversational AI: voice- and chat-based agents with low-latency requirements, built on modern telephony and real-time media infrastructure. LLM orchestration: agentic systems that pursue goals over multi-turn conversations, with hard behavioral guarantees enforced in code — not just prompts. Multi-channel session management: stateful user interactions that move seamlessly across channels. Retrieval and grounding: connecting LLM systems to large internal knowledge assets, with structured data flowing back. Evaluation infrastructure: measuring conversational quality against human baselines and gating rollouts on metrics, not vibes. Own services end-to-end: design, implementation, deployment (Kubernetes/Azure), and operations, including latency-sensitive real-time workloads. Design and tune LLM pipelines: context engineering, structured outputs, model selection, cost/latency tradeoffs across vendors. Build the eval loop and treat quality as a number: define metrics, run experiments, gate releases on them. Integrate third-party AI and voice vendors behind clean abstractions so providers stay swappable. Partner with product and compliance stakeholders to encode domain rules into the system. Set engineering patterns for a growing team: code review, observability, incident habits. What You Have: 6+ years of software engineering, with 2+ years building LLM-powered products in production (not notebooks): orchestration, RAG/grounding, structured outputs, eval-driven iteration. Strong TypeScript or Python; comfortable working across both. Real-time or streaming systems experience: WebSockets, WebRTC, audio pipelines, or comparable low-latency work. Cloud-native production experience: containers, Kubernetes, CI/CD, observability. Azure a plus. Familiarity with agent/LLM frameworks such as LangGraph, LangChain, or CrewAI — hands-on exposure is enough; we care that you know the landscape and when plain code beats a framework. Working familiarity with vector databases and embedding-based retrieval. Pragmatic product judgment: you scope prototypes, kill weak approaches with data, and ship. Nice-to-have: Voice AI: telephony platforms, conversational agent frameworks, STT/TTS, turn-taking/VAD. Knowledge graphs or large-scale search/retrieval systems. Regulated or compliance-sensitive domains (healthcare, finance). Eval…