AI/ML Engineer

rockITdata•Arlington, VA

17h•Remote

About The Position

Driven by Innovation and built on Trust, rockITdata is a unique SDVOSB services company that partners with leading commercial healthcare/life sciences organizations on cutting edge innovations - think AI, automation and data transformation. We then bring those commercially tested solutions to government entities to deliver predictable, measurable impact for the American taxpayer and consumer. rockITdata is seeking an AI/ML Engineer. This role is responsible for designing, building, and optimizing AI-powered capabilities within an Amazon Connect contact center environment running in AWS GovCloud (IL4). This role will focus on Amazon Bedrock Agent configuration, prompt engineering, knowledge base development, coaching ruleset design, QA scoring rubric implementation, and continuous model evaluation. The engineer will work across all four AI capabilities: IVR Call Taking, AI Coaching, Call Summaries, and Post Call Analysis.

Requirements

3+ years of experience in applied AI/ML engineering, with hands-on work building LLM-based applications (prompt engineering, RAG architectures, agent frameworks)
Direct experience with Amazon Bedrock, including Agent configuration, Knowledge Base setup, Guardrails, and foundation model invocation
Proficiency with Amazon Nova model family or equivalent foundation models (e.g., Claude, Titan) in production environments
Strong Python and/or Node.js skills for Lambda function development and AI pipeline integration
Experience with retrieval-augmented generation (RAG) patterns, including vector store management, chunking strategies, and semantic search optimization
Familiarity with AWS serverless services: Lambda, Kinesis, S3, CloudWatch, and IAM within GovCloud environments
Understanding of healthcare data privacy requirements (HIPAA, PHI/PII handling) and federal compliance frameworks (FedRAMP, NIST 800-53)
Experience with version-controlled prompt management and A/B testing of prompt variants
Strong analytical skills for interpreting model performance metrics and translating findings into prompt or architecture improvements
Bachelor’s degree in Computer Science, Data Science, AI/ML, or related field (or equivalent experience)

Nice To Haves

Experience with Amazon Connect, Contact Lens, and contact center AI workflows
Prior work in FedRAMP High or IL4/IL5 environments
Familiarity with Salesforce integration patterns (Platform Events, Streaming API, REST/SOAP APIs)
AWS certifications: Machine Learning Specialty, Solutions Architect, or DevOps Professional
Experience with TRICARE, DHA, or healthcare insurance domain knowledge
Background in conversational AI design (IVR, chatbots, voice agents)

Responsibilities

Configure and optimize Amazon Bedrock Agents for virtual agent fulfillment, real-time coaching, call summarization, and post-call quality analysis
Design, version-control, and iteratively refine prompt templates for each AI capability, measuring performance against SLA accuracy targets (≥95% QA match)
Build and maintain the AWS Bedrock Knowledge Base, including content ingestion pipelines synced from Salesforce and semantic retrieval optimization using Amazon Nova models
Develop coaching rulesets that map required disclosures, HIPAA-sensitive language patterns, and DHA-mandated scripts to specific call types for real-time compliance monitoring
Implement QA scoring rubrics for the post-call analysis platform, enabling automated evaluation of sentiment, compliance, accuracy, and resolution effectiveness across 100% of calls
Configure Bedrock Guardrails for PHI/PII redaction, hallucination detection, prompt-injection protection, and content filtering on every inference call
Evaluate new Amazon Nova model releases (Nova Pro, Nova Lite, Nova Micro) against production benchmarks for accuracy, latency, and cost efficiency, proposing model swaps through change management
Collaborate with the Integration Engineer to ensure Bedrock inference outputs are properly formatted for delivery to Salesforce (via Lambda and Platform Events) and to Verint
Monitor model performance using CloudWatch metrics and QuickSight dashboards; identify and remediate model drift at Day 30 and Day 60 re-benchmarking intervals
Participate in biweekly sprint demos, presenting AI performance data, prompt tuning results, and accuracy metrics to TriWest stakeholders
Work with TriWest QA reviewers and SMEs to close knowledge gaps identified through post-call analysis feedback loops
Ensure all AI inference and logging complies with FedRAMP High, DFARS 252.204-7012, HIPAA, and NIST AI RMF requirements