End-to-End QA for AI Marketing Platform

We are looking for an experienced QA engineer (or small QA team) to run complete end-to-end testing of Salfars (salfars.com), an AI marketing agent platform for ecommerce brands. The platform helps users run ad account audits, plan campaigns and generate ad creative (images/video) across Meta, Google, TikTok, Amazon Ads and Shopify, with human approval required before anything launches or spends. PREFERRED CANDIDATES We strongly prefer candidates or agencies with an in-house social media / performance marketing team. Such teams already understand ad platforms, can set up and manage test accounts correctly, and can judge whether the AI's audits, plans and ad copy are actually good marketing, not just technically working. MUST-HAVE REQUIREMENT: ACCOUNTS ON EVERY PLATFORM This project requires connecting Salfars to multiple platforms. You must already have, or be able to create within the first 2-3 days, your own accounts on each of the following: - Meta: Business Manager / Business Portfolio with a new Ad Account and a Facebook Page (Instagram account linked, if possible) - Google: a new Google Ads account (a manager/agency account with a separate test client account is ideal) - TikTok: a TikTok Business Center with a new TikTok Ads account - Amazon: an Amazon Ads account linked to a Seller Central or Vendor account (or tell us upfront if you cannot get this, so we can plan around it) - Shopify: a free Shopify Partner account with a development store, or a trial store with sample products Proposals that cannot provide accounts on all platforms will not be considered, unless you clearly state which platforms you cannot cover. TEST ACCOUNT SAFETY RULES (IMPORTANT) - Create NEW, dedicated test or agency accounts for this project. Do NOT connect any existing, live or client ad accounts, Business Managers, Pages or stores. We must not put any production account at risk of restrictions, disapprovals or billing issues. - Use accounts that belong to a real, legitimate business or agency identity. No fake identities, borrowed or purchased accounts, or any workaround that violates platform policies. - No real ad spend. Do not run live campaigns. Use draft or paused campaigns, minimal-permission roles, and budgets set to zero or kept in draft. Any step that would launch or spend must stop at the approval screen and be reported. - Keep account activity natural and low volume, so the new accounts are not flagged by the platforms. - Use dedicated test emails/logins (not personal or client credentials), and share credentials only through a secure password manager. Never in chat or email. - Use sample or dummy products, images and ad copy only. No real customer data. - Test with different permission levels (admin, standard, limited access) and also check disconnect/reconnect and expired-token behaviour. - Note: Salfars is in beta, and platform connections depend on provider approval and account permissions. Report clearly any connection that fails because of access limits, so we can separate platform issues from product bugs. SCOPE 1. UI / functional testing - Marketing site: home, solution pages, guides, pricing, about, careers, legal pages, links, forms, SEO meta tags - Sign-up / login / onboarding flow (app.salfars.com) and workspace creation - Pricing and plans: Free, Starter ($19), Growth ($49), Enterprise ($299); usage limits for images/video; upgrade, downgrade, cancel - Platform connection flows (Meta, Google, TikTok, Amazon, Shopify): connect, permission screens, errors, disconnect, reconnect, multiple accounts per platform - Chat copilot UI, approval/review screens, generated image and video outputs - Cross-browser (Chrome, Safari, Firefox, Edge), mobile responsive, basic accessibility and performance checks 2. AI agent evaluation (workflow testing) - Core workflows: AI ad account audit, AI campaign planning, AI ad creative - Cross-platform scenarios: the same brief across two or more connected platforms, and plans that combine Shopify data with ad platform data - Check agent outputs for accuracy, grounding in the connected account's real data, hallucinations and unsupported ad claims - Verify guardrails: no launch, spend or budget change without explicit user approval; correct account scope and permissions; the agent never touches an account that was not connected for the test - Test edge cases: empty/new accounts, incomplete briefs, conflicting instructions, prompt-injection attempts, long conversations - Score outputs against a simple rubric (accuracy, relevance, safety, marketing quality, usefulness) and track consistency over repeated runs DELIVERABLES - Test plan, test case suite and a platform/account setup checklist - Bug reports with steps to reproduce, severity, screenshots/screen recordings - Agent evaluation report with rubric scores and example good/bad outputs - Final summary report with a release-readiness recommendation - Automated regression tests (Playwright or Cypress) for key UI flows is a plus DAILY COMMUNICATION (required) - A daily 15-30 minute meeting (video call) to review feedback, bugs found, blockers and next day's plan - A short written daily summary of bugs reported and test progress - Availability overlapping with IST working hours REQUIREMENTS - Proven experience in end-to-end web application testing (manual + automation) - Experience testing AI/LLM-based products or agent workflows is strongly preferred - Hands-on experience with Meta, Google, TikTok and/or Amazon Ads and Shopify (in-house social media/performance marketing team preferred) - Familiarity with Jira/Linear/GitHub Issues, Postman, browser dev tools - Clear written English and good bug-reporting discipline TIMELINE: approx. 4 weeks, with an option to extend for regression cycles. Week 1 includes account setup and platform connection. Please include in your proposal: (1) which of the five platforms you can provide accounts for, (2) whether you have an in-house marketing team, (3) relevant AI/web testin