The short answer
ML Researcher focuses on advancing the field—publishing papers, running experiments on novel algorithms, and pushing boundaries. ML Scientist does research-adjacent work but prioritizes applied outcomes within a company. Applied Scientist is anchored to a specific business problem and ships production models. The same person may hold all three titles at different companies; the work type matters more than the label.
ML Researcher: Theory and Publication
An ML Researcher is typically tasked with advancing machine learning as a discipline. You're designing novel architectures, running large-scale experiments, and publishing results. Your success is measured by papers accepted at top conferences (NeurIPS, ICML, ICCV) or citations in the field.
This role often lives in:
- Big tech labs (Google Brain, Meta AI, OpenAI)
- Academic institutions
- Specialized AI labs (Anthropic, DeepMind)
You have more freedom to choose what to build, but less control over when it ships. The metric is novelty, not revenue impact. Many ML Researchers work independently or lead small teams of research scientists and engineers.
ML Scientist: Research With Guardrails
An ML Scientist does research work but reports to a business function. You experiment with new techniques and evaluate their potential, but the experiments usually target an internal business need—not field-wide advancement.
You might:
- Prototype new approaches to a company's recommendation system
- Run A/B tests on ranking algorithms
- Benchmark model architectures for a specific production constraint
- Author internal papers documenting findings
The outcome is often a decision—"should we use approach A or B for our search pipeline?"—rather than a published paper. You have more structure than an ML Researcher but more autonomy than an Applied Scientist. The metric is typically internal impact and velocity.
Applied Scientist: Problem-to-Production
An Applied Scientist owns a specific business outcome. You're handed a problem ("improve checkout fraud detection" or "reduce recommendation latency by 30%"), you build and ship a model, and you're accountable for the metric.
Your typical workflow:
- Scoping: Work with product, eng, and stakeholders to define success metrics
- Exploration: Run quick experiments and pick the most promising direction
- Validation: Test on holdout data and in production (A/B test)
- Shipping: Integrate into production and monitor
- Iteration: Refine based on live performance
You're deeply integrated with engineering. The metric is shipped impact—revenue lift, latency reduction, cost savings, whatever the business cares about. Many Applied Scientists spend half their time in code review, data pipeline debugging, or feature engineering rather than algorithm exploration.
How they differ in practice
| Dimension | ML Researcher | ML Scientist | Applied Scientist |
|---|---|---|---|
| Problem source | Self-chosen or PI-defined | Business or research leadership | Product or engineering team |
| Success metric | Publication, novelty, citations | Internal impact, research velocity | Shipped business metric (KPI) |
| Reporting | Research director, usually | Head of Research or Data Science | Head of Data Science, ML Eng, or Product |
| Time in production code | Minimal | Some prototyping, limited shipping | High—integration, testing, monitoring |
| Collaboration | Other researchers, grad students | Other researchers + business stakeholders | Engineers, PMs, Data Eng, Analytics |
| Publication expected? | Yes, strongly encouraged | Sometimes, internal focus | Rarely, NDAs often prohibit it |
Why the titles get confused
Big tech companies use these titles inconsistently. Google calls some roles "Research Scientist," Meta has "AI Researcher," and Microsoft uses "Principal Applied Scientist" for very senior roles. The same job description might appear under three different titles depending on the team and hiring manager.
When you see a posting, look at the reporting line, the stated OKRs, and the project examples. A role titled "ML Scientist" reporting to a product team and owning a KPI is really an Applied Scientist. A role titled "Applied Scientist" at a research lab reporting to a Director of Research might actually be an ML Researcher.
Which one should you target?
If you're in academia or transitioning from a PhD, ML Researcher roles let you continue publication-driven work with better funding and resources. The bar is high—these teams expect conference-ready novelty within your first year.
If you want research momentum but more stability and business context, ML Scientist is the middle ground. You're building things that matter to the company, but you're not on the hook for quarters-long production ownership. It's a common role for people bridging research and industry.
If you want to ship quickly, see impact, and work closely with engineers and product, Applied Scientist is the fastest path to tangible outcome. You'll spend less time on theory and more on debugging training pipelines at scale. The tradeoff: less time for long-horizon research and publication.
Get in front of the roles that fit
These titles often go live across multiple job boards within hours. Teams hiring for ML researchers, scientists, and applied scientists move fast—especially if they're backfilling or scaling an org. Using a platform like GiraffyReach that detects postings the moment they're live and auto-applies before the crowd floods in can be the difference between a callback and being lost in the noise.