Voice AI Agent

Sourcebae; India; Remote; INR 1,674,000 - 2,790,000; Full time

Job summary

Sourcebae is seeking an expert to own a production voice AI agent end-to-end from design to deployment. You will ensure stability, optimize transcripts, and drive value by transforming scripted flows into intelligent, low-latency conversations.

You will release through automated gates with PII/PHI safeguards and strict governance. You will work with Amazon Connect, Lex V2, Bedrock, Python, and cloud services to minimize latency and cost while handling sensitive data in a regulated industry.

Qualifications

• Built and launched at least one LLM app used by real customers or employees; strong in prompt engineering, tool calling, structured output, and retrieval-augmented generation.
• Hands-on with Amazon Connect and Amazon Lex V2: contact flows, Lambda integrations, intents and slots.
• Experience with Amazon Bedrock or comparable managed LLM service: knowledge bases, agents, guardrails.
• Strong Python skills and AWS core services: Lambda, IAM, DynamoDB, S3, CloudWatch, KMS.
• Experience with sensitive data in regulated industries under formal change control.

Responsibilities

• Own a production voice AI agent end-to-end, maintaining stability and transcript-driven tuning.
• Redesign scripted verification plus Q&A into a fully agentic, low-latency conversation.
• Ship releases through automated quality gates with PII/PHI safeguards and governed DevOps pipelines.
• Monitor performance and cost of voice AI agents and continuously optimize.

Skills

LLM app development

Prompt engineering & RAG

Python

Data handling in regulated industries

Tools

Amazon Connect

Amazon Lex V2

Amazon Bedrock

Lambda

DynamoDB

S3

CloudWatch

KMS

Job description

Duration - 6-month contract, possibly extended

Overlap until 9:30 PM IST

Location - Remote

Skills -

AWS Ecosystem for Voice AI Amazon Connect, Lex, Bedrock, Python, Eleven Labs, good to have Sales Force integration. This is for Voice AI Agents

Job Description

Own a production voice AI agent end to end: keep it stable through incident and transcript-driven tuning, redesign it from scripted verification plus Q&A into a fully agentic, lower-latency conversation, and ship every release through automated quality gates, PII/PHI safeguards, governed Azure DevOps pipelines, and active monitoring of performance and cost.

Must-have skills (please screen for these first)

• Has built and launched at least one LLM app used by real customers or employees. Demos and proofs of concept don't count. Strong in prompt engineering, tool calling, structured output, and RAG.
• Hands-on with Amazon Connect and Amazon Lex V2: contact flows, Lambda integrations, intents and slots
• Amazon Bedrock or a comparable managed LLM service: knowledge bases, agents, guardrails
• Strong Python, plus AWS core services: Lambda, IAM, DynamoDB, S3, CloudWatch, KMS
• Has worked with sensitive data in a regulated industry (insurance, healthcare, or financial services) under formal change control

Strongly preferred

• Voice AI experience, such as ElevenLabs or similar text-to-speech and speech-to-text tools, and reducing latency in real-time voice apps
• Agentic orchestration: Bedrock Agents or Flows, Step Functions, EventBridge, SQS/SNS
• Terraform, or CDK/CloudFormation
• Azure DevOps, including YAML pipelines with AWS service connections
• Java (Spring Boot) or TypeScript/Node

Nice to have

• Contact Lens, Amazon Q in Connect, Kinesis
• Vector stores such as OpenSearch Serverless or pgvector
• Insurance or benefits domain background

Not a fit

• Data scientists or ML researchers who focus on training or fine-tuning models
• People who have only built chatbots on no-code platforms
• People whose AI experience is limited to courses, hackathons, or personal projects