Founding Voice AI Engineer (Part-time, potential path to Chief AI Officer)

## About the role The work spans the AI side of Namecoach — wherever applied AI, speech AI, and pronunciation intelligence meet our product surface. The list below is illustrative. You'd weigh in on priorities with the founder, and the mix will shift as the company grows. ## What you'll do - Core pronunciation AI for Euphonia. Build the AI systems that learn from corrections and improve pronunciation predictions and controls. Concretely: fine-tune G2P and pronunciation models on our proprietary dataset (millions of verified audio pronunciations and phonetic spellings); build multi-model AI validation pipelines that produce confidence scores and model provenance; design data architecture for verified pronunciation entries that persist and reuse high-confidence results; build context-aware ranking systems that pick the right pronunciation given identity, locale, and prior confirmations; and develop related proprietary work we'll cover in the defense round. This is our flagship work and where you'd likely start. We'll go deeper in the take-home and the defense round. - Voice AI platform integrations. Build MCP-style integrations and SDKs so developers building on voice infrastructure and orchestration platforms can drop in Namecoach pronunciation quality with a few lines of code. - Enterprise customer work. Implementation work for current partners and customers, ranging from the Fortune 500 to famous sports organizations. Sometimes deeply technical (fine-tuned models on customer-specific name data, on-prem deployments), sometimes scrappy (one-off audits, hand-transcribing IPA for a high-stakes customer launch). - Pronunciation benchmarks. Voice AI lacks an authoritative measurement framework for pronunciation reliability across providers, locales, and word categories. We want our team to define one — see the Publish and position as an expert section below. - Reverse-mode pronunciation coaching. Same underlying tech, applied in reverse: a human attempts a name, the system gives per-phoneme feedback. Use cases include ceremony announcing, customer-facing role prep, language learning, and onboarding. - Internal AI tooling and agents. Help Namecoach itself operate as an AI-native company. We're small; we want someone who builds agents, automations, and internal tools that multiply everyone's leverage — operational dashboards, knowledge-graph RAG systems, growth-loop automations, and beyond. - Consumer product experiments. We have a small backlog of consumer-facing ideas (Namecoach for Individuals, name-search Chrome extension, etc.). If you want to ship a consumer-facing AI product or features for them, there's room. ## The 80/20 rule applies throughout Ship product 80% of the time, do targeted research and writing that builds individual and company expertise the other 20%. ## How we ship AI-assisted development is our default working style — Claude Code, Cursor, agent frameworks, and the modern agentic stack are the primary interfaces to writing code at Namecoach. This applies across everything we ship: internal tools, POCs, customer demos, and production product features alike. The CEO ships this way today, and we want you as well — ideally pushing the team's leverage further than we currently do. If you're already shipping with these tools as your primary development interface, you'll be at home. ## Get close to the data Pronunciation quality requires actually-close-to-the-data work — sometimes that means listening to dozens of audio clips, hand-transcribing IPA and Namecoach-format phonetics, or tracking down why a single customer's name set is broken. We need someone who finds this clarifying rather than tedious. Some weeks you'll be deep in customer-specific work; other weeks on core AI infrastructure. Both matter. People allergic to the messy parts of real-world data don't tend to succeed here. ## Publish and position as an expert Pronunciation reliability in voice AI is an emerging measurement problem — there's no agreed-upon benchmark or widely-used canonical evaluation framework. We want our team to be the authoritative voice on this, both because it's good for the field and because it's good for the company. We'd support you in: - Publishing technical blog posts on what's under our hood, how baseline systems compare (frontier LLMs and older TTS providers), and what we've learned from production - Authoring or co-authoring papers — benchmarking studies, applied-ML methods papers, position pieces on pronunciation reliability as a measurement problem - Speaking at industry conferences