Private Draft

The 29 personas behind AI

We’ve organized every stage and persona in the AI supply chain, informed by real recruiting at frontier companies. Click any row to see matching profiles from our talent graph.

Shaped by Industry Experts

← ATOMS & ENERGYUSERS & MARKETS →

← Back

Applied ML

Trains models for products

Applied ML

Summary

Known as: Applied Scientist, Applied Researcher, Applied Research Scientist, ML Engineer, Machine Learning Engineer

Bridge between research and product. Innovates on model capabilities for specific domains and ships them into production. The broadest ML hiring category by volume, spanning applied researchers who publish in service of product, production ML engineers who own models end-to-end against business metrics, and foundation ML engineers who build shared capabilities multiple product surfaces consume. At big tech, this is the dominant engineering function; at startups, it's often the entire ML team.

Specializations

Applied Research — Innovates on methods within a product-facing domain. Publishes, runs novel experiments, and pushes capability in service of a specific product surface. Targets known product needs and iterates against product metrics alongside research metrics. Examples: video and image generation (Runway, Midjourney, Stability), search and retrieval (Cohere, Perplexity), speech and audio synthesis (ElevenLabs, Suno), music generation (Spotify), multimodal understanding (Apple).

Production ML Engineering — Owns a model end-to-end in production: problem formulation, training, serving, and iteration against business metrics. The largest Applied ML sub-pool by volume, covering recommendations, ranking, content understanding, ad targeting, fraud, delivery optimization, and forecasting. At big tech (Meta, Google, Amazon, TikTok, Spotify, Snap) this is the dominant ML hiring category and directly revenue-linked.

Foundation & Shared Capabilities — Builds ML systems that multiple product surfaces consume: embeddings, retrieval, multimodal understanding, content understanding, and shared model infrastructure. Adapts and maintains shared model capabilities downstream of pre-training for internal consumers across product lines. Examples: Reality Labs Foundation (Meta), shared search/retrieval (Cohere, Perplexity).

The sub-pools (applied research, production ML, foundation/shared) have different interview loops, comp bands, and career trajectories. The domain axis — vision, speech, search/retrieval, recommendations, generative media — is the primary specialization that determines which teams a candidate is relevant to. At model-is-the-product companies (Runway, ElevenLabs, Midjourney), applied research closely resembles frontier research in technical depth — the distinction is the product mandate, not the sophistication. Scientific ML (healthcare, drug discovery, materials science) is a distinct sub-category where domain expertise constrains everything.

Where the Work Lives

[1]Substrate

[2]Compute

[3]Intelligence

Primary

Innovates on model capabilities for specific domains — vision, speech, recommendations, search.

[4]Systems

Primary

Ships domain-specific models into production with the reliability and latency requirements of real products.

[5]Distribution

Secondary

Model quality directly drives product metrics and revenue at companies like Meta, Google, and Amazon.