Founding Machine Learning Engineer

About Us

We are a VC-backed, stealth, agentic AI startup co-founded by Alex Dalyac (AI unicorn founder) and Alexandre de Brébisson (voice cloning inventor, PhD supervised by Yoshua Bengio), whose founding team includes staff engineers from GitHub, Amazon and an indie hacker with 1M+ app downloads.

We want to create an economy where professionals don't just use AI, they own AI; turning their expertise into income-generating AI agents.

Our iOS app and AI agents are starting to spread among professional circles, especially in healthcare, legal & home.

We're looking to bring on a world class Founding ML Engineer.

The Role

As a Founding Machine Learning Engineer, you will lead technical direction and implementation within our voice agents for communication & admin, from rapidly deploying the state-of-the-art to pioneering novel approaches in constitutional AI and meta-learning. You’ll have a large degree of autonomy and responsibility, with an expectation to work across the stack to not only research but also embed your proprietary AI features into the product; making them reliable, discoverable, intuitive & useful.

Our current stack is V-FiST-G-PoP: Vapi, Firebase, Swift, TypeScript, GCP, PostgreSQL, Python (FastAPI, SQLAlchemy).

Key Responsibilities:

• Implement state-of-the-art: natural voice AI (expressive, turn detection, multilingual), agentic AI for communication & admin (eg calendar, calls, texts, emails, paperwork), prompt templates & eval sets, within an iOS app
• Pioneer novel approaches in constitutional AI, meta-learning and continual learning, where the agent autonomously refines its custom prompt for each user. The agent learns on the job by evaluating how well it did according to user’s objectives & feedback, modifies its prompt to do better next time, and makes sure its new prompt is optimal by evaluating it on past and simulated scenarios
• One foot in applied AI, the other in engineering: not just shipping AI models, but full stack AI product features that are reliable, discoverable, intuitive & useful to users
• Promote a culture of being constantly at the cutting edge of AI breakthroughs, productizing them faster than the market, with an eye to developing proprietary AI edge

Relevant literature:

• Constitutional AI: Harmlessness from AI Feedback, where instead of scoring responses as good/bad (RLHF), the AI is trained to align its behavior with a set of principles or a "constitution." This constitution can be updated based on high-level feedback
• Reflexion: Language Agents with Verbal Reinforcement Learning, where agent learns from trial & error by reflecting on its own "verbal feedback" (a textual critique of its past actions)
• Large Language Models as Optimizers, where an LLM "engineers" prompts for another LLM, using "Optimization by PROmpting" (OPRO)

We’re Looking For:

• 5+ years of professional experience as ML engineer or AI scientist
• LLMs, voice AI, agentic AI, prompt template design, evals, data ops
• LLM-as-judge, constitutional AI, meta-learning, continual learning, prompt learning from verbal feedback, speech-text synthetic data
• Personal/Executive AI assistant domain tasks
• Expertise productizing AI breakthroughs, building upon them & developing a proprietary edge
• Top tier ML publications are a plus; strong scientific understanding of ML literature is a must
• Fluency in our full tech stack is a big plus; full stack software engineering skills & pride are a must
• Open source contributions in AI are a major plus
• Experience in an early stage startup or other fast-moving high responsibility/autonomy environment
• Potential for relocation to NYC down the line, or to found an applied research hub where you’re currently based

What We Offer:

• Founding member and shape the future of a product you can call your own
• Competitive salary and equity options
• Flexible working environment with potential to relocate to NYC or found an applied research hub where you’re currently based
• A culture that values collaboration, creativity, and continuous improvement