Senior Agentic Software Engineer

We're looking for a Senior Engineer who is a true self starter- someone who sees an ambiguous problem space, defines the path forward, and ships. Rather than waiting for tickets, this person will identify what needs to be built, make the call, and drive it to production. This role sits at the center of Pearl's agentic product strategy. You'll be building autonomous systems that reason, act, and integrate, across the dental practice- work that defines what AI can do in clinical operations, not just what it can describe., * Agentic Systems (primary focus): Design and build agents that plan, call tools, and take action across internal services and external systems. This includes tool-calling workflows, ReAct-style loops, MCP-based integrations, evaluation harnesses, and the safety guardrails that make autonomous behavior trustworthy in a clinical context.

• Real-Time Clinical AI Workflows: Contribute to systems that capture clinical audio, stream transcription, and generate structured documentation using LLMs.
• Dental Data Intelligence: Help develop pipelines that ingest, normalize, and enrich data from practice management systems - the substrate our agents reason over.

What You'll Do

• Architect and build agentic systems: tool-calling loops, orchestration patterns, memory, retries, and evaluation
• Define and enforce safety guardrails for autonomous workflows operating on clinical and operational data
• Integrate agents with internal Pearl services and third-party tools via MCP and direct APIs
• Build evaluation pipelines and observability for agent behavior - measure what's actually working in production
• Contribute to real-time transcription and LLM-powered summarization pipelines
• Develop across the stack: Node.js/TypeScript APIs, React frontends, PostgreSQL, and AWS
• Support event-driven architectures using Kafka, SQS, or similar technologies
• Own prompt engineering, structured output design, and model selection decisions for your domain
• Identify gaps in the system before anyone files a ticket, then close them
• Work with AI development tools (e.g., Cursor, Claude Code, Copilot) as a core part of your workflow

• 5+ years of professional software engineering experience
• Self-starter by default. You operate well without detailed direction. You scope your own work, make tradeoff decisions independently, and deliver - then bring the result back for discussion rather than waiting for permission to begin.
• Comfort with ambiguity. You can take a vague problem ("our agent should handle this class of request") and turn it into a working system without a detailed spec
• Hands-on experience building LLM-based systems: prompting, structured outputs, tool/function calling
• Experience designing or contributing to agentic patterns - ReAct, tool-calling loops, multi-step reasoning, or similar
• Strong backend skills with Node.js / TypeScript
• Experience with React and modern frontend development
• Familiarity with AWS (Lambda, SQS, S3, Bedrock preferred)
• Experience with PostgreSQL and data modeling fundamentals
• Exposure to event-driven systems (Kafka, SQS, etc.)
• A real testing and observability practice - agentic systems fail in subtle ways and you know how to catch it
• Daily use of AI-assisted development tools in your workflow

Nice to Have

• Experience shipping production agentic systems (not just prototypes)
• Familiarity with MCP (Model Context Protocol) or comparable tool integration frameworks
• Experience with agent evaluation frameworks (LangSmith, Braintrust, custom eval harnesses)
• Experience with speech-to-text systems (Deepgram, Whisper, AWS Transcribe)
• Healthcare or dental technology background
• ETL pipeline or data integration experience
• Multi-tenant SaaS experience
• Infrastructure as Code experience (e.g., Serverless Framework, CDK)

• Competitive compensation and benefits
• Ongoing training and development opportunities
• Flexible, uncapped PTO
• Remote work environment