Upcoming Event: The AI Agent Implementation Framework
Upcoming Event: The AI Agent Implementation Framework
EnterOne
Reston, Virginia
IT Training & Professional Services
Amazon Bedrock (Claude Haiku 4.5, Amazon Titan Embed Text v2), AWS Lambda, Amazon API Gateway, Amazon OpenSearch Serverless, Amazon S3, Amazon SQS, Amazon RDS, Amazon CloudFront, AWS CloudFormation, Amazon CloudWatch, AWS IAM, AWS WAF, AWS Secrets Manager
EnterOne, a veteran-owned IT training and professional services company serving customers in over 35 countries, partnered with Avahi to build an intelligent, conversational AI agent that augments course discovery on its training platform. Avahi delivered “Nia,” a multi-agent chatbot built on Amazon Bedrock that lets users find the right IT training course through natural conversation rather than keyword filtering across hundreds of pages. The solution will be live in production on EnterOne`s website, embedded into EnterOne’s Ruby on Rails application, and will serve both authenticated and anonymous users with contextual recommendations across EnterOne’s IT training catalogue spanning Cisco, VMware, Microsoft, AWS, ITIL, and other vendors.
Founded by engineers for engineers, EnterOne provides online and in-person IT training and consulting services to professionals and enterprises navigating fast-moving networking, cloud, and AI technology shifts. Headquartered in Reston, Virginia, with offices in Tyner, North Carolina and Latin America, the company trains over 10,000 professionals worldwide and delivers more than 200 IT professional services projects each year. EnterOne’s training catalogue spans associate-level certifications through expert tracks across major vendor ecosystems, complemented by custom training, advanced services, labs, and software development offerings.
EnterOne’s existing website search relied on keyword matching and faceted filters across a large, vendor-segmented catalogue. Prospective students typically had to know exactly what they were looking for, click through multiple category pages, and manually compare schedules, prerequisites, and certification paths before booking. For users earlier in their journey, those exploring a new certification, comparing learning paths, or unsure which course matched their experience level, this friction translated directly into abandoned sessions and lost bookings.
The training market is increasingly competitive, with learners expecting conversational, AI-augmented discovery experiences similar to those they encounter on consumer platforms. Without modernizing the search experience, EnterOne risked falling behind on user expectations and missing enrollment opportunities, particularly from professionals whose intent was clear but whose vocabulary did not match the catalogue’s structured filters (for example, asking about “container training” when the relevant offerings were Docker and Kubernetes courses).
EnterOne needed a solution that could understand intent, hold a conversation, filter the live course database by real-world criteria like date, cost, and format, and surface course content details, all while integrating cleanly with the existing Rails-based website and class management system.
EnterOne selected AWS as the cloud platform for its existing infrastructure footprint and for the breadth of services required to build a production-grade, secure, and cost-efficient generative AI system. Amazon Bedrock provided managed access to Anthropic’s Claude foundation models with no model hosting overhead, while Lambda, API Gateway, OpenSearch Serverless, S3, and SQS enabled a fully serverless architecture that scales with demand and avoids idle infrastructure cost.
The AWS ecosystem also met EnterOne’s security and operational requirements out of the box: IAM for fine-grained access control, AWS WAF and VPC isolation for the public API surface, CloudFormation for infrastructure-as-code, and CloudWatch for observability across every component of the pipeline.
EnterOne selected Avahi based on Avahi’s standing as an AWS Partner with deep, hands-on expertise across the AWS AI/ML stack, particularly in building production agentic systems on Amazon Bedrock. EnterOne needed a partner who could move beyond proof-of-concept demos and deliver a hardened, observable, maintainable system ready to embed in a live customer-facing website.
Avahi’s track record of designing multi-agent architectures, vectorization pipelines, and serverless backends, combined with disciplined engineering practices around evaluation, prompt versioning, and infrastructure-as-code, gave EnterOne confidence that the engagement would result in a system its team could own, extend, and trust in production.
Avahi designed and built a multi-agent conversational AI system, branded “Nia,” composed of two complementary AWS pipelines and a React-based chat widget designed to embed directly into EnterOne’s Ruby on Rails application.
The synchronous Conversational AI Pipeline handles real-time user interactions. User queries flow through Amazon API Gateway (protected by AWS WAF) into the Course Agent Lambda function, which orchestrates a five-agent architecture powered by Claude Haiku 4.5 on Amazon Bedrock. An Orchestrator agent analyzes user intent and routes work to specialized agents in parallel: a DB Agent that translates natural language into safe, read-only SQL queries against EnterOne’s RDS database (returning live availability, pricing, schedules, and instructor data); a RAG Agent that performs semantic search across vectorized course materials in Amazon OpenSearch Serverless; and a Follow-up Agent that handles clarifying questions and multi-turn context. A Synthesizer agent then composes the final response, verifies course details against the live database, and injects programmatic course URLs before returning to the user.
The asynchronous Vector Search Enhancement Pipeline keeps the knowledge base current. When new course PDFs are uploaded to S3 or new courses are added to the database, an Event Handler Lambda enqueues vectorization jobs to Amazon SQS. A Vectorize Lambda consumes the queue, extracts content from course PDFs, generates embeddings using Amazon Titan Embed Text v2, and indexes them into OpenSearch Serverless. A dead-letter queue captures failures for inspection, and the pipeline scales automatically with content volume.
Session management is handled by a lightweight S3-backed store that persists conversation history for both authenticated EnterOne users and anonymous visitors. This allows Nia to maintain conversational continuity across page navigations and return visits without requiring login.
The frontend is a Create React App-based chat widget built in TypeScript, served from an S3 bucket fronted by Amazon CloudFront, and designed to mount into a Rails-controlled DOM node within application.html.slim. Once deployed to production, the widget will load on every page of http://enterone.com , share authentication context with the host application, and communicate with the backend through a REST API exposed by API Gateway.
The entire infrastructure is defined in AWS CloudFormation templates and deployed via a GitHub Actions CI/CD pipeline. Avahi also conducted a comparative evaluation of Claude Haiku 4.5 and Claude Sonnet 4.5 using a COMET-inspired semantic similarity framework, ultimately selecting Claude Haiku 4.5 for its superior balance of response quality, speed, and cost-efficiency.
The Nia AI agent is production-ready and prepared for deployment on http://enterone.com , giving EnterOne’s prospective students a conversational, intent-aware path to course discovery that will complement the existing website search. Evaluation results show that Nia successfully handles a wide range of query types, from precise lookups (“show me courses starting December 15th”) to open-ended exploration (“I want to get AWS certified but don’t know where to start”), while gracefully escalating to human support when queries fall outside its scope. EnterOne now has a maintainable, observable, and extensible AI platform that its team can evolve independently once live, with documented procedures for adding new vendors, certifications, filters, and even entire specialized agents as the business grows.
Let’s explore your high-impact AI opportunities together in a complimentary session
