高级 QA 自动化工程师

We are looking for an experienced Senior QA Engineer to own quality for our AI voice virtual agent. This role is for someone who is equally comfortable reasoning about a large language model's behavior and reading a SIP trace.
You should be interested in testing non-deterministic systems. A voice agent takes a live phone call, transcribes speech that may be accented or noisy, interprets intent, reasons over an LLM, generates a spoken reply, and decides whether to resolve the request or hand it to a human—all within a latency budget callers will notice. Nothing in that chain can be validated with a simple assertion of equality, so you will design the evaluation approach that makes it testable anyway.
You should also be interested in the underlying telephony. Testing this agent means placing real calls: provisioning SIP endpoints and softphones, driving calls programmatically, controlling media conditions, and distinguishing a model failure from a codec, jitter, or call-routing issue.

• PLEASE SUBMIT CV IN ENGLISH.

• ONLY ACCEPTING CANDIDATES FROM LATAM.

• ADVANCED SPOKEN ENGLISH IS REQUIRED.

Responsibilities

• Embed fully within a Scrum team, participating in all ceremonies and maintaining associated artifacts.

• Own the end-to-end test strategy for the AI voice virtual agent across the LLM, speech, and telephony layers.

• Design test scenarios for complex conversational behavior, including multi-turn dialogues, interruptions and barge-in, ambiguous or incomplete input, out-of-scope requests, human escalation/transfer, and failure recovery.

• Build and maintain an evaluation framework for non-deterministic output, leveraging golden datasets, semantic/rubric-based assertions, and automated regression detection for prompt, model, or retrieval changes.

• Adversarially test the agent against hallucinations, incorrect grounding, guardrail bypasses, prompt injections, and unauthorized disclosure of customer data.

• Prepare and provision SIP devices, endpoints, and call scenarios to automate call origination end-to-end.

• Diagnose failures across the entire stack, isolating issues down to the LLM, ASR (speech recognition), TTS (text-to-speech), media path, or call routing layer.

• Extend automated coverage and integrate test suites into the CI/CD pipeline to serve as release gates.

• Define and track core quality metrics, including task completion, containment, intent accuracy, transcription accuracy, response latency, and escalation correctness.

• Maintain documented test coverage in the TMS with full traceability to business requirements.

• Communicate and escalate quality risks to stakeholders backed by data and evidence.

Requirements:

• 5+ years in QA, with recent hands-on ownership of a complex, integrated product.

• Practical experience testing SIP/VoIP systems, including call flows, SIP signaling, media handling, softphones, or SIP trunking, with the ability to troubleshoot using SIP traces or packet captures.

• Demonstrated creativity in test design (live scenario design will be a core part of the interview process and heavily weighted).

• Proficiency in Python, sufficient to independently build and maintain automation frameworks.

• Experience testing non-deterministic or probabilistic systems, with a clear approach to designing assertions for variable outputs.

• Strong analytical skills, including the ability to define quality metrics, justify their calculation, and use data to back release decisions.

• Hands-on experience with formal TMS tools and Jira.

• Professional fluency in English (spoken and written), with reliable overlap with US Eastern business hours.

We use the following technologies:

• Python with pytest for test automation.

• Playwright for browser and WebRTC-based call driving.

• JsSIP / WebRTC for programmatic call origination and media control.

• SIP softphones and devices for manual and exploratory testing.

• Allure TestOps as our TMS, linking automated runs to documented test cases.

• Jira for project management and issue tracking.

• Jenkins & GitHub Actions for CI/CD pipelines.

• Grafana for quality and delivery metrics.

• LLM APIs and evaluation tooling — currently in active development, and we are open to any effective approach.

Pluses:

• Direct experience testing LLM-based or conversational AI products, including evaluation frameworks and LLM-as-a-judge approaches.

• Familiarity with speech technologies, including ASR/transcription accuracy measurement (WER), TTS quality assessment, endpointing, and silence detection.

• Experience with RAG or retrieval-grounded systems, specifically in evaluating grounding correctness and context retrieval.

• WebRTC and media-layer testing experience, including codec behavior and performance degradation under jitter or packet loss.

• Background in testing IVR, contact center, or conversational IVR platforms.

• Multilingual testing capabilities, with emphasis on Spanish/English code-switching.

• Literacy in prompt engineering, sufficient to analyze system prompts and identify potential failure modes

We are pleased to offer:

• Remote work opportunity!

• B2B Employment ($, gross).

• Stable job with long-term growth perspective.

• Competitive salary with annual performance review.

• Really good hardware.

• An exciting and challenging job with talented people around.

• Continuous learning and career growth opportunities.

• Compensation for professional training, seminars, and conferences.

• Referral program: get rewarded for helping us grow the team with talented people.

• Company-supported English classes to enhance your professional growth.

Creativity is a core requirement for this role, not a nice-to-have. The defects that matter here are found by someone who imagines the caller who interrupts mid-sentence, switches languages halfway through, gives a partial account number, answers a different question, or deliberately tries to jailbreak the agent. Scenario design is the primary skill we are hiring for.

Current State: We have automated call testing built on JsSIP/WebRTC with Python and Playwright, alongside a growing body of test scenarios. The evaluation layer—how we score responses consistently, catch regressions when a prompt or model changes, and report quality metrics leadership can act on—is where this role will have the most impact.

About us:

IDT Technologies is a part of IDT Corporation—an American telecommunications company founded in 1990 and headquartered in New Jersey. It is an industry leader in prepaid communication and payment services and one of the world's largest international voice carriers. We are listed on the NYSE, employ over 1,500 people across 20+ countries, and have revenues in excess of $1.5 billion.

Net2Phone () is a global provider of unified cloud communications solutions. It offers an integrated communications suite featuring powerful collaboration tools, including voice, messaging, chat, and unlimited domestic and international calling to 23+ countries.

PLEASE SUBMIT CV IN ENGLISH

Originally posted on Himalayas