Founding ML Engineer

ABOUTThe Persona Labs is building a new kind of social platform focused on something most social products do not explicitly optimize for: helping people become real friends.We want to help people discover interesting people around them, find meaningful common ground, start low-pressure interactions, continue promising conversations, create shared experiences, and ultimately build real-life friendships.DISCOVER → CURIOSITY → COMPATIBILITY → INTERACTION → UNDERSTAND → IRL EXPERIENCE → FRIENDSHIPTHE AI LAYER - COMPANION INTELLIGENCE Alongside the platform, we are building a proactive personal AI companion that learns about the user and helps them navigate this journey through personalized recommendations, suggestions, reminders, conversations, and experiences. THE OPPORTUNITY We are looking for a Founding ML Engineer to build the intelligence layer of the platform from the ground up. This is a 0→1 Applied AI / ML role where you will work directly with the founder and Product Engineer to turn ambiguous problems around users, relationships, recommendations and personal intelligence into working systems. You will be expected to: Understand the problem → identify the signals → design the intelligence system → prototype → evaluate → deploy → learn → improve.WHAT YOU WILL BUILD & OWN USER INTELLIGENCE User representations, behavioural models, interests, preferences, contextual signals, and evolving understanding of the user. MEMORY Short- and long-term memory, episodic/preference/relationship memory, retrieval, relevance and updating. RECOMMENDATION & MATCHING People discovery, compatibility, activity/experience recommendations, and personalized ranking. INTENT & INTEREST Infer what the user is trying to do and learn what they care about from behaviour, not only declared interests. RANKING Decide what should appear first across potentially thousands of relevant people, activities or experiences. CONTENT INTELLIGENCE Classification, toxicity, spam, policy signals, quality, relevance, and semantic understanding. RELATIONSHIP INTELLIGENCE Reciprocity, interaction health, shared interests, progression, declining engagement and shared activity. NEXT-BEST-ACTION Determine the most useful action now: show a person, suggest a question, recommend an activity, reconnect, or do nothing. TRUST / SAFETY INTELLIGENCE Fake-account signals, spam, abuse, behavioural anomalies, risky interactions and moderation assistance. COMPANION INTELLIGENCE Use signals and outputs to help the companion decide what to say, suggest, recommend or not do. WHAT YOUR DAY-TO-DAY LOOKS LIKE • Translate ambiguous product problems into ML/AI system designs. • Build models and intelligence pipelines using behavioural, relational and contextual signals. • Develop recommendation, matching and personalization systems. • Design memory and retrieval systems that help the companion understand the user over time. • Build and evaluate LLM-powered and agentic workflows. • Decide when to use traditional ML, rules, retrieval, ranking or LLMs. • Prototype quickly, test assumptions and iterate based on real user behaviour. • Work closely with the founder and Product Engineer to turn intelligence into product experiences. • Design APIs and production systems that bring ML/AI capabilities into the application. • Build evaluation, monitoring and feedback loops so the intelligence improves over time.WHO SHOULD APPLY • Experience: 0–4 years’ experience, including exceptional fresh graduates. Strong 1–3 year engineers and experienced 3–4 year product builders are welcome. • Strong foundations in ML, Python, statistics and software engineering. • Evidence of Building: Experience with AI/ML projects, recommendation systems, LLM applications or personalization is highly valued. • Strong evidence of building: Shipped projects, research, hackathons, internships, open source or startup work. WHAT WE LOOK FOR MACHINE LEARNING DEPTH Can you understand the modelling problem underneath the application? RECOMMENDATION & PERSONALIZATION Can you reason about relevance, ranking, cold start and behavioural signals? AI ENGINEERING Can you turn LLMs and agents into reliable product capabilities rather than simple API wrappers? USER INTELLIGENCE Can you design systems that gradually understand a person from sparse and changing signals? SYSTEMS THINKING Can you move from a model to a production system with APIs, data, latency, cost and monitoring? EVALUATION MINDSET Can you determine whether the intelligence actually helped the user? PRODUCT JUDGMENT Can you decide what the system should do when there is no predefined answer? SPEED OF EXECUTION Can you move from idea → prototype → evaluation → production quickly and responsibly?BUILD WITH US You will join at a stage where many of the answers do not exist yet. You will not simply implement a model someone else selected; you will help decide how the product learns to understand people.CAREERS:Apply with your resume, GitHub, portfolio or shipped work.[hidden]