Backend/AI Engineer

Acaisoft is a company partnering with industry leaders across healthcare, telecom, fintech, and insurance, building Cloud, Data & AI solutions for fast-moving businesses. Although based in Poland, most clients are in the U.S., requiring flexibility around non-standard working hours for smooth collaboration across time zones. The role focuses on developing the AI Voice Dispatch Assistant, a scalable, production-ready multi-agent voice AI solution running entirely within the client's AWS perimeter.

• Availability to maintain partial overlap with the Mountain Time (MT) zone
• Minimum 5 years of experience in Python engineering
• Practice in developing high-performance, resilient backend services and APIs using FastAPI
• Strong practical knowledge of AI/ML concepts, especially LLM orchestration, Retrieval-Augmented Generation (RAG), and prompt engineering
• Expertise in AWS services including ECS/Fargate, Lambda, API Gateway, DynamoDB, ElastiCache, and Bedrock services (Nova 2 Sonic, Nova Micro)
• Familiarity with the Strands Agents SDK or other multi-agent architecture frameworks

AI Agent Development:

• Design, implement, and maintain the multi-agent system using the Strands Agents SDK (BidiAgent) on ECS/Fargate
• Ensure per-session isolation, auto-scaling, and manage the Nova 2 Sonic BidirectionalStream lifecycle

Core Backend Services:

• Develop and maintain the production FastAPI backend
• Include a centralized AHS API client library with features such as Pydantic models for complex responses
• Implement per-call audit logging with field-level PII masking and per-endpoint caching

LLM Orchestration & RAG:

• Implement the triage super-agent for first-turn domain routing (Nova Micro) and delegate to specialized sub-agents
• Integrate Bedrock Knowledge Bases (OpenSearch Serverless, Titan embeddings) via RAG wiring to ground agent responses in technical manuals

Security and Compliance:

• Configure and wire Bedrock Guardrails for PII detection, prompt-injection defense, and content-moderation thresholds
• Implement authentication using AWS Cognito, FastAPI JWT middleware, and custom Lambda authorizers

Infrastructure & Deployment:

• Implement and manage serverless components (API Gateway, Python Lambda handlers) for service request creation and STREEM escalation triggers
• Manage session state using DynamoDB and ElastiCache Redis

Availability to maintain partial overlap with the Mountain Time (MT) zone is required.