Miju Labs

← All buyers · Luma AI

S

Sabrina Day

Evaluation (co-author, Ray3 Evaluation Report) · Luma AI

low confidenceTier A · Talk to firstImage, video & audio modelsvideo

Day is one of two named authors of Luma's Ray3 Evaluation Report (October 2025), which is the most detailed public description by any video lab of a human-rated creative evaluation framework. She and Simran Motwani define the seven dimensions Luma grades its video models on and state that outputs were rated by trained evaluators with diverse domain expertise. Almost nothing else about her is public, so she is high-value and low-visibility at the same time.

Background

Public evidence for Day consists of the Ray3 report byline and a LinkedIn profile listing her at Luma AI. No prior employers, education, talks or other publications were found; a similarly named academic site (sabrinafielder.com) appears to be a different person and should not be attributed to her.

Notably, she does not appear on the user-maintained Org listing of Luma's HQ staff, which lists Simran Motwani as a software engineer, so her internal title is unconfirmed.

What they run now

The Ray3 evaluation framework. The report grades video on physics and motion performance, instruction-following, motion artifacts, aesthetic quality, dynamic range and colour reproduction, temporal consistency, and image-to-video adherence, across what it calls thousands of generated videos spanning diverse prompts, styles and difficulty levels. It describes the framework as the product of cross-disciplinary collaboration between scientists, filmmakers, artists and creators, and references independent review and third-party quantitative evaluation without naming any vendor or panel.

What they buy

Strong inference. Somebody recruited, trained and managed the expert evaluators behind that report, and Luma's careers page currently lists no roles for evaluation, annotation or human feedback among 32 openings — which means that work is either done by a very small internal team or bought. The addressable purchases are: recruited professional rater panels (filmmakers, colourists, VFX), rubric development and inter-rater-reliability work, and independently run comparative evaluations Luma can cite. No vendor is named in the report, which is itself the opening: an evaluation report that claims independent review without naming the reviewer has a credibility gap.

Career

  1. c. 2025 – presentEvaluation — co-author, Ray3 Evaluation Report, Luma AI — Title inferred from the report byline and LinkedIn; not confirmed by Luma

On the record

who should rate video · 2025-10

States that Ray3 outputs were rated by trained evaluators with diverse domain expertise against standardised success metrics. source

framework provenance · 2025-10

Credits the evaluation framework to collaboration between scientists, filmmakers, artists and creators. source

Recent posts

Sabrina Day
Product Evals Lead @ Luma AI | Multimodal & Agentic AI | Expressive Tech | ex-Meta, Oculus
"AI creators are not replacing great creatives. Great creatives are becoming dramatically more powerful." 👏
171 reposts

Interviews & talks

Why it matters here

Top three in this batch on fit, even though the dossier is thin. She holds the exact job function Julian's company exists to serve, at a company that publishes its rubric and therefore has something to defend. Luma also has European presence — a Forward Deployed Creative role posted in Germany and a London office — which makes a German vendor less exotic than it would be for a pure-US lab.

How to reach

Speculation, but the best-shaped opening in the batch: write to her on LinkedIn about the report itself, specifically the two soft points — that 'trained evaluators with diverse domain expertise' is unfalsifiable without stated qualifications and counts, and that 'independent review' is not independent if it is internal. Offer to run one dimension of the framework (motion artifacts, or aesthetic quality) as a genuinely external panel with published rater credentials, sample sizes and inter-rater agreement, on Ray 3.2 against FLUX 3 — which BFL has just claimed beats Ray 3.2 in 93% of comparisons. That comparison is a live commercial wound for Luma.

What we could not establish
  • Her actual job title, tenure and background — none found.
  • Whether Luma used an external vendor for the Ray3 raters.
  • Whether she is still at Luma (she is absent from the user-maintained Org roster).

Why they are on the map

Co-authored the Ray3 human evaluation report (trained expert evaluators, creative-industry framework). LinkedIn lists her at Luma AI. source — evidence dated 2025-10

Runs vendor projects; decides if a pilot converts.

How Luma AI buys data

Ray3 Evaluation Report (2025-10-14, by Sabrina Day and Simran Motwani) says outputs were 'rated by trained evaluators with diverse domain expertise' and that the framework was built with 'scientists, filmmakers, artists, and creators' plus 'independent review'. The vendor is not named. 2026 hiring: RL researcher to 'build reward infrastructure… reward-model serving, LLM-as-judge', a World Model researcher to 'define and own the metrics', and a PM for Applied Research. Leadership stresses curated, quality-over-quantity data.

Founder-led (CEO Amit Jain, CTO Alex Yu) with Chief Scientist Jiaming Song over research. Visible teams: Foundation Models, Training Compute (RL), Data Infrastructure, Inference, Forward Deployed Creative, and Dream Lab LA. Evaluation reports come from a small team (Day, Motwani).

Sources

  1. Ray3 Evaluation Report — State-of-the-Art Performance for Pro Video Generation (Luma AI, Oct 2025)
  2. Sabrina Day — LinkedIn
  3. Luma AI careers
  4. Luma AI — HQ team listing (The Org; user-maintained, treat as weak evidence)