Member of Technical Staff - Recommendation System Engineer

Member of Technical Staff - Recommendation System Engineer

San Francisco Bay Area

·

Full-time

level dependent

Build the shared personalization layer that decides what Nirva should surface, for whom, and when. An engineering role for someone who connects memory, ranking, and feedback to make journals, insights, and conversations more useful and personal with every day of use.

← BACK TO CAREERS

Member of Technical Staff - Recommendation System Engineer

San Francisco Bay Area

·

Full-time

·

180k-350k

·

level dependent

ABOUT NIRVA

Nirva is building the first AI fashion accessory for your mind — a necklace or bracelet that privately listens only to the wearer to enable proactive journaling, emotion tracking, and life coaching. Think Oura Ring for your Mind.

Backed by Lightspeed and South Park Commons, recognized at CES within 90 days of founding, and launching in the U.S. this fall. Co-founders bring backgrounds from Meta Reality Labs (Quest, Ray-Ban glasses) and fashion industry.

THE ROLE

We’re looking for a Recommendation Engineer to build the system layer that connects Nirva’s memory, context, and user understanding to every content generation feature we ship, and makes all of them better. Today each generated experience — a journal entry, a daily recap, an emotional insight, a proactive nudge, a conversation starter, a chat response — can end up rebuilding its own retrieval, personalization, and timing logic. You will build the shared layer underneath them: what to surface, for this person, at this moment, in what form, and why. The result is a product that does not just generate more content, it generates the right content, at the right time, and gets more personal with every day of use. You will work directly with the founding team to build this layer from the ground up.

WHAT YOU'LL OWN

• Build the recommendation and personalization layer that sits between personal context / memory and all content generation features: journaling, daily recaps, emotional insights, proactive experiences, coaching suggestions, conversation starters, and chat

• Design ranking and decisioning: relevance, recency, emotional salience, novelty, diversity, and usefulness, plus timing, frequency, and when to say nothing at all

• Build a unified signals and context service for generation features, so product surfaces consume ranked, explained personal context through one layer instead of each rebuilding retrieval and personalization

• Build feedback loops: implicit and explicit user signals, dismissals, edits, and engagement, turned into ranking improvements, personal preferences, and data flywheels for evals and model improvement

• Own cold start and learning over time: how Nirva is useful on day one with almost no data, and measurably more personal by week four

• Build offline and online evals for generated content: personal relevance, specificity, repetition, timing, and trust, with A/B testing and failure analysis across features, not just click-through

• Work across LLMs, embeddings, retrieval, and ranking to improve generated output quality: context selection, prompt inputs, structured outputs, and post-generation ranking and filtering

• Partner with product, design, memory, and audio teams to make every generated surface feel warm, specific, trustworthy, and emotionally intelligent, not generic or gimmicky

WHAT WE'RE LOOKING FOR

Must-haves

• 2+ years of full-time engineering experience, or a PhD in computer science, machine learning, information retrieval, recommender systems, or a related field, plus 2 applied internships (industry or applied research) where you shipped or evaluated real systems

• Hands-on experience building recommendation, ranking, search, or personalization systems used by real users, with clear evidence of impact on quality or engagement

• Strong working knowledge of retrieval, embeddings, ranking, and context engineering for LLMs, and how to connect a ranking layer to content generation, not just to a feed

• Experience with feedback loops and experimentation: logging the right signals, offline evals, A/B tests, and separating “users clicked it” from “users actually found it useful and personal”

• Strong debugging and eval mindset for AI systems: you can identify why a generated output felt generic, mistimed, or repetitive, and fix the system, not just the prompt

• Strong product taste for consumer products that need to feel personal, trustworthy, and emotionally resonant, and strong product engineering instincts overall

• High ownership: you identify what needs to happen and drive it to shipped quality, moving quickly across ambiguous AI, product, and engineering problems without waiting for perfect specs

• Energized by 0-to-1 work, fast iteration, and early-stage startup intensity

• Based in San Francisco and able to work in person

Strong signals (nice-to-haves)

• Built personalization for a consumer AI, companionship, journaling, coaching, wellness, or content product

• Experience with proactive / ambient products: notifications, nudges, digests, or timing-sensitive surfaces where restraint matters as much as relevance

• Memory systems, personal knowledge graphs, long-term user modeling, or context stores

• LLM eval frameworks, human feedback loops, or fine-tuning for personalization

• Experience spanning backend, ranking infrastructure, and user-facing surfaces, including mobile

COMPENSATION

San Francisco Bay Area · Full-time · 180k-350k, level dependent

TO APPLY

Apply here