Senior Automation QA (JavaScript) Engineer

Ciklum is looking for a Senior Automation QA (JavaScript) Engineer to join our team full-time in India.

We are a custom product engineering company that supports both multinational organizations and scaling startups to solve their most complex business challenges. With a global team of over 4,000 highly skilled developers, consultants, analysts and product owners, we engineer technology that redefines industries and shapes the way people live.

About the role

As a Senior Automation QA (JavaScript) Engineer, become a part of a cross-functional development team engineering experiences of tomorrow. Ciklum is seeking a Senior AI Automation QA Engineer to join the client project, an ambitious initiative to build an enterprise-grade Agent Development Platform (ADP) that our client will commercialize to their own customers. This is quality engineering for a system that is non-deterministic by nature, where "expected output" is a statistical claim rather than a string match, and where the platform's reliability guarantees no lost workflow runs, strict tenant isolation, graceful degradation under provider outages are contractual commitments, not aspirations. You will build the automation that proves those guarantees hold, release after release.

You will own the test automation that spans the whole platform: the Frontend and Backend that agent developers depend on, the long-running Temporal workflows that must survive worker crashes and provider failures, the multi-tenant isolation boundaries that protect one customer's data from another, and the performance and concurrency targets defined in the platform's NFRs. Much of your craft is adversarial and failure-first injecting worker kills, retry storms, and malformed documents to verify the system recovers exactly as specified, and negative-testing disallowed tool calls and egress paths to confirm the sandbox holds. Where agent behavior must be judged rather than asserted, you will partner closely with the evaluation function, contributing the test infrastructure, synthetic-data pipelines, and CI integration that make quality measurable and repeatable.

This is a high-visibility role where we establish the production-ready foundation the rest of the program is built on. Your automated suites become release gates, and the evidence they produce feeds directly into client milestone sign-offs so your work must be reproducible, trustworthy, and defensible to a sophisticated client. As the platform scales toward an "agent-a-month" delivery cadence, you will build the reusable test scaffolds and harnesses that let quality keep pace with delivery, ensuring that month twelve ships as safely as month one. You will work shoulder-to-shoulder with AI engineers, platform engineers, and the evaluation team, embedding quality from the design stage rather than inspecting for it at the end.

Responsibilities

Design, build, and maintain automated test suites for the platform, Frontend and Backend, including contract tests that protect the agent-developer experience as the SDK evolves

Build automated tests for long-running Temporal workflows including failure injection (worker kill, provider outage, retry storms) to verify durable-execution guarantees such as zero lost runs and correct compensation/saga behavior

Automate multi-tenant isolation verification cross-tenant data access, configuration bleed, and cost-attribution correctness executed per release as a platform acceptance criterion

Verify agent execution sandboxing and egress controls through negative testing of disallowed tool calls and network destinations

AI-Specific & Non-Deterministic Testing

Design test strategies for LLM-driven behavior: statistical assertions over repeated runs (pass^k consistency), semantic-similarity scoring, confidence-threshold validation, and flakiness quarantine that distinguishes model variance from genuine regression

Build and continuously extend adversarial test corpora document-borne prompt injection in lease PDFs, tool-call hijack, and data-exfiltration probes aligned to OWASP LLM Top 10 and recognized red-team taxonomies

Verify Human-in-the-Loop (HITL) escalation behavior confidence thresholds fire correctly and no gated/irreversible action completes without authorization

Partner with the evaluation function on golden-dataset-driven checks, contributing test infrastructure, synthetic-data (synthetic lease) pipelines, and CI integration that make agent quality measurable and repeatable

Performance, Reliability & NFR Verification

Automate verification of platform NFRs invocation overhead (p95), concurrency targets, queue backpressure behavior, and LLM provider failover time and surface trends per release

Use tracing data (Langfuse / OpenTelemetry spans) as a first-class test asset: assert on trace completeness, token/cost-accounting accuracy, and alerting behavior for anomalous runs

Process, Evidence & Collaboration

Wire test suites into CI/CD as hard release gates with automated, per-metric regression flagging on pull requests

Produce auditable, reproducible test-evidence packs that feed directly into client milestone sign-off

Embed with AI and platform engineers from the design stage making acceptance criteria, testability, and observability inputs to architecture, not afterthoughts

Maintain test environments, mocked LLM/provider layers, and synthetic data generators that keep test runs fast, deterministic where possible, and affordable

Mentor mid-level QA engineers, codify AI testing patterns into reusable internal frameworks, and contribute to Ciklum's quality engineering practice and AI Academy

Requirements

8+ years of test automation with API testing, and building test frameworks other engineers adopt, not just test cases

Hands-on experience testing LLM-powered or other non-deterministic systems statistical assertions, semantic scoring, and managing model variance (this is the differentiator; cite real examples)

Experience testing as