Miju Labs

← All buyers · Google DeepMind

Nicole Brichtova

Director of Product, Visual Generation (Nano Banana, Imagen, Veo) · Google DeepMind

high confidenceTier A · Talk to firstFrontier labsimage · video

Product lead for Google DeepMind's entire generative-media line — Nano Banana / Gemini Image, Imagen, Veo and Omni. She has said publicly, and with visible discomfort, that modelling teams end up setting the world's default aesthetic and are not obviously the right people to do it, and she has named the data she cannot manufacture internally: real-world professional task trajectories. That is the single most explicit buying signal from any frontier-lab visual-product lead in this batch.

Background

Brichtova came to generative media from strategy rather than research: Georgetown, then Duke's Fuqua School of Business, then Deloitte consulting and Google consumer product strategy before moving into DeepMind's product organisation. She is described in the 36Kr profile of the Nano Banana team as Head of Visual Generation Products, and in the August 2026 AI Engineer panel as Director of Product for generative media models.

Her public account of how Nano Banana was built is unusually data-centric for a PM. She attributes the character-consistency breakthrough to obsessive data curation rather than scale, describes building evaluation around team members' own faces because you cannot judge likeness on strangers, and describes the 'infographic explosion' as a default-aesthetic problem: the model was too cluttered and too eager to show everything it knew, and had to be actively tuned back. She also describes a community-testing layer — artists at Google and DeepMind, and executives, playing with the models — to capture what benchmarks miss.

What they run now

Product for Google DeepMind's visual generation family: Nano Banana Pro (Gemini 3 Pro Image), Nano Banana 2 (Gemini 3.1 Flash Image) and Lite, Imagen, Veo and the Omni line. Her job includes deciding the model's default aesthetic — saturation, palette, density — and deciding what evaluation the team trusts when benchmarks cannot separate two candidates.

What they buy

The strongest evidenced demand in this batch, though no vendor is named. She has stated the requirement in public: real-world professional task trajectories — e.g. the full path of a marketing campaign from a product photograph through multiple ad formats — which cannot be synthesised internally. Around that, her practice implies: curated high-quality reference data for specific capabilities (likeness, text rendering), large-scale human evaluation for subjective quality, artist-panel community testing, and aesthetic tuning judgements on defaults such as density and saturation. The buyable products are trajectory capture from working creative teams, artist panels with structured rubrics, and default-aesthetic preference studies across cultures and audiences.

Career

  1. —Consultant, Deloitte
  2. —Consumer product strategy, Google
  3. presentProduct Manager → Group PM → Director of Product, generative media, Google DeepMind — Nano Banana / Gemini Image, Imagen, Veo, Omni
  4. —Georgetown University, Bachelor's
  5. —Duke University, Fuqua School of Business, MBA

On the record

Who sets default aesthetics · 2026-08

Modelling teams end up setting the world's default aesthetic, and they are not obviously the right people to do so. source

Nano Banana's default look · 2026-08

The model's default was too cluttered and too eager to show everything it knew, and had to be tuned back. source

What they cannot make internally · 2026-08

She asks for real-world task trajectories — what a marketing campaign actually looks like from product photo to multi-format ad assets — which are hard to manufacture internally. source

Data quality over scale

Character consistency required obsessive attention to data quality, not just scale. source

Evaluating likeness

It is very difficult to judge character consistency on faces you do not know, so the team evaluated using its own members' faces. source

Human evaluation

They run community testing with artists at Google and DeepMind and with executives, because benchmarks miss the emotional and qualitative dimension. source

Interviews & talks

Why it matters here

Highest in this batch. She is a product leader with budget, an explicit unmet data need she has described publicly, and an intellectually honest position that the aesthetic defaults should not be set by modelling teams alone — which is the argument for buying outside taste rather than relying on internal instinct. Julian's entire thesis is a direct answer to a request she has already made on a podcast.

How to reach

Quote her need back to her with a worked example. Speculation on channel, not on substance: a short, concrete artefact — one real marketing campaign trajectory, captured end to end from a working agency (brief, product photo, concepts, rejected directions with reasons, final multi-format assets), with the rationale annotated — plus a note saying 'you said these are hard to manufacture internally; here is one, and here is what a hundred would cost and look like'. Send via her X account or LinkedIn. The rejected directions with reasons are the part no internal team can produce and no scraper can find.

What we could not establish
  • Exact current title — 'Group Product Manager', 'Director of Product' and 'Head of Visual Generation Products' all appear across sources; Director of Product is the most recent (Aug 2026).
  • Whether Google DeepMind procures creative data externally or only via Google-internal artist programmes.
  • X handle confirmed from an X profile listing, not from a bio naming the employer.
  • Date of the Sequoia episode not captured.

Why they are on the map

Product lead for Nano Banana; describes eval stack of 'large-scale human evals for subjective quality', trusted testers and artist reviews; Imagen 3 core contributor source — evidence dated 2025-10

Needs the data for a model; writes the spec.

How Google DeepMind buys data

Imagen 3 report (2024-08) collected 366,569 ratings from 3,225 raters in 71 countries via an unnamed 'external platform' under GDM data-enrichment policy (living wage). Veo 3.1 page cites human side-by-sides on MovieGenBench (1,003 prompts), VBench I2V and internal sets. Nano Banana leads (2025) describe large-scale human evals plus trusted testers and artist reviews; Gemini Data team job post (2026) covers data curation/generation/evaluation.

Aug 2026: Kavukcuoglu runs GDM incl. Gemini as SVP; Hassabis chair. Gemini has Data, Post-Training, Safety Post-Training and Model Behavior teams (job posts); Gemini app sits inside GDM under Josh Woodward. Gen-media (Imagen/Nano Banana/Veo/Flow) has product lead Brichtova and modality research leads (Wang, Erhan).

Sources

  1. How Google's Nano Banana Achieved Breakthrough Character Consistency
  2. Leads of Nano Banana, Imagen, Veo, Omni/Thinking (AI Engineer panel)
  3. Uncover the Team Behind Google's Viral Nano Banana
  4. Google DeepMind Developers: How Nano Banana Was Made
  5. Gemini Image — Nano Banana (model page)