Product Manager - Voice AI

About ClearGrid ClearGrid is an AI-native debt resolution platform operating across the UAE and Saudi Arabia. Our AI voice agents handle borrower conversations at scale in Arabic and English for leading banks and NBFIs, running on infrastructure we built in-house. This is one of the largest production deployments of conversational voice AI in the region. Real regulated clients, real volume, and real consequences when a call goes wrong. We're past the demo stage and deep into the harder problem: making voice AI reliable, affordable, and scalable. About The Role We're hiring a Product Manager to own our Voice AI product end to end. This is a high-autonomy, high-visibility role for someone equally comfortable debating voice model budgets and model trade-offs with engineers as they are shaping roadmap and working with commercial teams and bank clients. You'll own three things that are currently unowned: the performance and cost of the voice stack, the tooling our prompt engineers depend on every day, and the evolution of the platform into something clients can configure and monitor themselves. You will be the single owner of Voice AI at ClearGrid, its quality, its economics, and where it goes next. What You'll Own Voice AI Stack Performance Own end-to-end performance of the voice product: quality, latency, transcription accuracy (Arabic and Gulf dialects especially), interruption handling, and conversational reliability. Work with engineering to instrument, monitor, and improve it continuously. Voice Infrastructure & Capacity Own the stack: STT, TTS, model routing, telephony gateways, dialing and redial capacity. Forecast capacity against volume so constraints surface in planning, not in production. Cost & Efficiency Own the unit economics of voice: cost per minute, cost per connected call. Build a defensible model, track it, and drive it down without sacrificing quality. Make build/buy and vendor decisions with data behind them, and find ways to serve a wider range of client budgets. Prompt & Agent Platform Own the tooling our prompt engineers depend on. Today, every AI agent across lenders, products, delinquency stages, and languages is authored, versioned, tested and deployed by hand. You'll turn that into a product: a central agent library, version history and rollback, automated pre-deployment validation, and regression testing against real call scenarios. The measure is simple: how fast we can launch a new agent, and how fast we can fix a live one. Production Quality Own how we detect and prevent failures in live agents. Our clients are regulated financial institutions; a bad call is a compliance event, not a bug. Build the QA and monitoring loop that catches issues systematically, rather than relying on someone listening to calls after the fact. Client-Facing Platform Lead the evolution of the voice platform into a product lenders can use directly: agent setup and configuration, campaign controls, performance analytics, call recordings and transcripts, QA evidence. Reduce the manual work that currently sits between a client request and a client answer. Product Discovery Track the voice AI market closely. Evaluate new models and capabilities, decide which languages and models we support, and know when something better has shipped. Stakeholder Ownership Own intake, prioritization, and delivery for everything voice. Work daily with prompt engineers, engineering, analytics, QA, compliance, and the collections floor, and be willing to say no to protect the roadmap. What We're Looking For • 4–6 years of product management experience, shipping software with engineering teams. • Hands-on voice AI experience: voice agents, speech models, or voice-based products: strongly preferred. We'll also seriously consider strong conversational AI / LLM application, contact centre, or speech backgrounds where the technical depth is clearly there. • Experience owning an internal tooling or workflow product for expert users. You'll be building for our own prompt engineers, and that's a distinct skill from client-facing product work. • Comfortable owning a cost line. You can build and defend a unit economics model and drive cost per unit down without degrading quality. • Data-literate. You pull and interrogate your own data rather than waiting for a dashboard. • A technical foundation (computer science or similar) is a plus, though hands-on experience matters more than formal qualifications. • Experience shipping Arabic or other non-English language products is a strong plus. Arabic fluency is a plus but not required; English fluency is. • Experience with regulated or enterprise clients, financial services especially, is a plus. • A systems thinker: hands-on, proactive, and comfortable operating with ambiguity. • An excellent communicator who can flex between an engineering standup and a bank steering committee. • Genuine enthusiasm for AI and emerging technology: someone who stays current on the space out of curiosity, not obligation. What Success Looks Like In 3 months: You've mapped the voice stack and the prompt engineering workflow, established baselines for deploy time, fix time, and cost per connected call, and agreed a Phase 1 roadmap. In 6 months: Manual prompt engineering effort is measurably down. A cost-per-connected-call reduction plan is documented and in execution. Client-facing setup and analytics are live with at least one lender. Show more Show less

4–6 years of product management experience, shipping software with engineering teams. Hands-on voice AI experience: voice agents, speech models, or voice-based products. Experience owning an internal tooling or workflow product for expert users. Comfortable owning a cost line and building/defending a unit economics model. Data-literate; you pull and interrogate your own data. A technical foundation is a plus, though hands-on experience matters more. Experience shipping Arabic or other non-English language products is a strong plus. Experience with regulated or enterprise clients, financial services especially, is a plus. Systems thinker with hands-on approach and comfort operating with ambiguity. Excellent communicator capable of switching between engineering standups and bank steering committees.

Own end-to-end performance of the voice product: quality, latency, transcription accuracy (Arabic and Gulf dialects), interruption handling, and conversational reliability. Instrument, monitor, and improve the voice stack with engineering. Own the voice infrastructure and capacity: STT, TTS, model routing, telephony gateways, dialing and redial capacity; forecast capacity against volume. Own unit economics: cost per minute, cost per connected call; build a defensible model and drive cost down without degrading quality; make build/buy and vendor decisions with data. Own the tooling for prompt engineers: central agent library, version history/rollback, automated pre-deployment validation, regression testing; measure speed of launching and fixing agents. Own Production Quality: build QA and monitoring loops to prevent regulatory/compliance issues in live calls. Lead Client-Facing Platform evolution: agent setup, campaign controls, analytics, call transcripts, QA evidence; reduce manual steps between requests and responses. Product Discovery: monitor voice AI market, evaluate models/languages, decide support scope. Stakeholder Ownership: intake, prioritization, and delivery for voice; coordinate with prompt engineers, engineering, analytics, QA, compliance, and collections; know when to say no to protect the roadmap.