Agentic AI Engineer

About this role

Taxbit is hiring an Agentic AI Engineer to design and deploy autonomous AI systems that solve complex tax and compliance workflows for enterprise customers. You'll leverage AWS agentic frameworks to build production-grade AI products that operate reliably in high-stakes financial contexts, working across the full lifecycle from prompt engineering through monitoring and iteration.

What you'll do

• Design and deploy customer-facing agentic AI systems using AWS services (Bedrock, Strands Agents, AgentCore, Lambda)
• Architect multi-agent systems with reliable tool use, memory management, and orchestration for financial accuracy
• Own full AI feature lifecycle including prompt engineering, evaluation, deployment, and production monitoring
• Collaborate with product, customer success, and compliance teams to translate user problems into AI solutions
• Write production-grade, well-tested code and conduct code reviews to maintain quality standards
• Monitor agentic systems in production and iterate based on user feedback and observability data

What they're looking for

• Python and/or TypeScript
• AWS AI/ML services (Bedrock, Strands, AgentCore, SageMaker)
• LLM agentic patterns (tool use, RAG, memory, multi-agent orchestration)
• Prompt engineering and LLM pipeline evaluation
• AWS cloud infrastructure (Lambda, IAM, networking)
• AI observability and monitoring tools
• Production software engineering and code quality practices
• Ability to communicate complex AI concepts across technical and non-technical audiences

Benefits

• Competitive salary: $130,000–$170,000 plus equity (RSUs)
• Comprehensive benefits package with 401(k) featuring crypto and financial wellness options
• Hybrid work: 3 days in-office, 2 days flexible/remote
• Discretionary unlimited time off plus paid parental leave
• Fertility benefits and modern workplace perks (team lunches, snacks)
• Autonomous work environment with flexibility in how you perform your role

Likely interview questions

• Describe a production LLM application you built—what challenges did you face with reliability, hallucination, or accuracy, and how did you address them?
• Walk us through how you'd design a multi-agent system to handle a complex tax compliance workflow. What patterns would you use for tool use and memory?