What is Shashwat Prakash's area of expertise?

Shashwat Prakash is an AWS Conversational AI & Contact Center AI Engineer specializing in enterprise Voice AI, IVR Automation, Amazon Connect, Generative AI applications, RAG Pipelines, AI Agent systems with Tool Calling, Nova Sonic real-time voice AI, and Prompt Engineering. He has 6+ years of experience delivering production-grade AI solutions for enterprise clients.

Which AWS services does Shashwat Prakash specialize in?

Shashwat Prakash specializes in Amazon Connect (IVR & Contact Center), AWS Bedrock (Generative AI & LLM), Amazon Lex (NLP & Chatbots), Nova Sonic (Real-time Voice AI), Amazon Q in Connect, AWS AgentCore (MCP Server), AWS Lambda (Serverless), AWS DynamoDB, AWS API Gateway, AWS S3, AWS CloudFront, AWS Cognito, and AWS EventBridge.

What is Shashwat Prakash's experience with IVR and Voice AI?

Shashwat Prakash has extensive experience as an IVR AI Developer and Voice AI Engineer, building production Amazon Connect IVR flows, Nova Sonic real-time conversational voice AI, AI voicebots, Speech-to-Text pipelines, Text-to-Speech systems, and full contact center automation solutions for enterprise clients at PwC US and Genpact India.

Does Shashwat Prakash offer Generative AI and RAG Pipeline development?

Yes. Shashwat Prakash is an experienced Generative AI Engineer and RAG Pipeline Developer. He builds retrieval-augmented generation systems, vector database integrations, semantic search, LLM-powered applications, Prompt Engineering frameworks, and Agentic AI systems with Tool Calling using AWS Bedrock and other foundation models.

Is Shashwat Prakash available for hire as a Conversational AI consultant?

Yes. Shashwat Prakash is available as a Conversational AI Consultant and AWS Contact Center AI Engineer for enterprise projects. He can be reached through the contact form at shashwatprakash.com for consulting, collaboration, and full-time opportunities.

How does a Voice AI agent integrate with Oracle Cerner EHR?

The voice AI agent connects to Oracle Cerner via its FHIR R4 REST API. When a patient calls, Amazon Connect routes the call to Amazon Lex, which captures the patient's intent. AWS Lambda then makes authenticated FHIR API calls to Cerner — fetching available appointment slots, patient demographics, and provider schedules — and returns the data to Lex in real time. All PHI is transmitted over TLS and stored encrypted using AWS KMS.

How was call latency reduced from 8–12 seconds to 3–4 seconds?

Three optimisations drove the latency reduction: (1) Parallel tool-calling — instead of sequential Lambda invocations, the agent executes multiple FHIR API calls concurrently; (2) Strategic model selection — switching to a lower-latency foundation model on Amazon Bedrock for intent classification; and (3) Prompt engineering — restructuring prompts to reduce token count without losing context, cutting inference time significantly.

What AWS services are used to build a HIPAA-compliant healthcare voice bot?

The solution uses Amazon Connect (HIPAA-eligible telephony), Amazon Lex (conversational AI), AWS Lambda (business logic), Amazon Bedrock (LLM inference), AWS KMS (encryption at rest and in transit), Amazon CloudWatch (audit logging), and Amazon S3 (call recording and log archival). All services are configured within a HIPAA-eligible AWS Business Associate Agreement (BAA).

Can patients book, reschedule, and cancel appointments through a phone bot?

Yes. The AI agent handles the complete appointment lifecycle over a natural voice call. Patients can speak their request — 'I want to book an appointment with Dr Smith next Tuesday' — and the agent queries Cerner for availability, confirms the slot, and writes the booking back to the EHR. The same flow supports rescheduling and cancellation, all without a human agent.

What is Nova Sonic and why was it chosen for this project?

Nova Sonic is Amazon's speech-to-speech foundation model available on Amazon Bedrock. Unlike traditional TTS + STT pipelines, Nova Sonic processes audio end-to-end, dramatically reducing latency and producing more natural-sounding responses. It was chosen for this project specifically because sub-4-second response times are critical for a conversational phone experience in healthcare.

PwC USOct 2025 – PresentAWS Conversational AI & Contact Center Engineer2 min read

Healthcare Conversational AI Agent with Oracle Cerner Integration

End-to-end voice AI that lets patients book, reschedule, and cancel appointments over a natural phone call.

The Business Problem

A major healthcare provider needed patients to self-manage appointments without waiting in a phone queue or navigating a complex web portal. Existing IVR menus were rigid, couldn't access live EHR data, and required a live agent for any schedule change.

Solution & Technical Overview

Built an end-to-end voice AI agent for a healthcare provider that enables patients to manage appointments through natural phone conversations. Architected on AWS using Amazon Connect (telephony), Amazon Lex with Nova Sonic (conversational engine), AWS Lambda (business logic), and Amazon Q in Connect (response generation), with secure FHIR-based integration to Oracle Cerner EHR for real-time patient data and provider schedules.

The agent handles the complete appointment lifecycle — booking, rescheduling, cancellation, and viewing. All patient health information (PHI) is encrypted at rest and in transit using AWS KMS, with comprehensive audit logging via CloudWatch and S3.

The primary technical challenge was end-to-end call latency. Through parallel prompt engineering, strategic LLM model selection, and redesigning the tool-calling architecture for concurrent API execution, response time was reduced from 8–12 seconds to a consistent 3–4 seconds — resulting in a natural, human-like conversational experience.

Key Outcomes & Results

Reduced call response latency from 8–12 s to 3–4 s via concurrent tool-calling architecture
Secure FHIR integration with Oracle Cerner EHR for real-time schedule access
PHI encrypted at rest and in transit with AWS KMS; full audit trail in CloudWatch & S3
Handles full appointment lifecycle: book, reschedule, cancel, and view
Nova Sonic (speech-to-speech model) for natural, low-latency voice interaction

Technologies & Skills Used

Amazon ConnectAmazon LexAWS LambdaAmazon BedrockNova SonicAmazon Q in ConnectAgentCoreAWS KMSOracle Cerner / FHIRPythonPrompt Engineering

Frequently Asked Questions

Written by

Shashwat Prakash

AWS Conversational AI & Generative AI Developer with 6+ years building enterprise contact center solutions, voice AI agents, RAG pipelines, and AI-powered applications on AWS. Currently at PwC US, previously at Genpact and Agenty Analytics.

LinkedIn GitHub Book a Free Consultation

Amazon ConnectAmazon Lex