Applied Scientist - Multimodal Guardrails & Evaluation

The success of generative AI guardrails is not determined solely by model architecture; it is defined by the quality of the concepts we protect, the robustness of our detection science, and the realism of our evaluation methodology . The Adobe Firefly Applied Science & Machine Learning team is expanding its IP guardrail systems for building s afe and compliant image , video, and audio gener ative models . A central challenge is long-tail IP coverage: foundation models cannot internalize every protected concept.

Advancing our guardrails therefore requires principled approaches to multimodal knowledge integration, detection modeling, and real-world evaluation science. We are seeking an Applied Scientist to drive the scientific development of IP concept modeling, detection strategies, and evaluation frameworks that determine the robustness ceiling of our guardrail systems. This is a research-oriented role focused on advancing multimodal detection and evaluation methodologies at production scale.

Research Areas You Will Drive Multimodal Concept Modeling & Knowledge Integration Develop principled methods for representing and organizing large-scale IP concept spaces across text, image, and audio. Study how retrieval-augmented systems (RAG), embedding alignment, and structured knowledge can complement multimodal foundation models. Investigate strategies for improving long-tail concept coverage beyond what VLMs inherently encode.

Design concept modeling techniques that meaningfully influence downstream guardrail decisions. Data Acquisition & Curation for Detection Develop scalable approaches for acquiring and curating high-quality multimodal datasets that improve detection coverage for IP-sensitive concepts. Drive long- tail expansion through targeted data collection, web-scale sourcing, and synthetic data generation.

Analyze failure cases in generative outputs to inform targeted data acquisition and dataset refinement. Design efficient data curation and labeling strategies that improve signal quality and robustness of downstream detection systems. Evaluation Methodology & Benchmark Design Define evaluation frameworks that reflect real-world Firefly usage patterns.

Design multimodal benchmark datasets that stress-test guardrails under realistic and adversarial scenarios. Develop metrics that capture over-blocking, under-blocking, semantic similarity, and near-miss generation. Establish statistically rigorous offline and online evaluation strategies that guide research prioritization.

Study how evaluation quality constrains and enables system-level progress. Feedback-Driven Scientific Iteration Leverage large-scale product feedback signals to identify systematic weaknesses in guardrail behavior. Translate real-world interaction patterns into structured evaluation hypotheses.

Build reproducible experimental pipelines that enable continuous scientific iteration. What You’ll Do Lead scientific design of IP concept expansion and detection methodologies. Formulate hypotheses around long-tail coverage, detection robustness, and evaluation gaps.

Run rigorous experiments to quantify performance ceilings and identify high-leverage improvements. Partner closely with generative model scientists to ensure alignment between detection, guidance, and evaluation systems. Contribute to intellectual property and potential publications in multimodal learning, evaluation science, or AI safety.

What You’ll Need to Succeed Research & Technical Depth PhD or MS in Computer Science, Machine Learning, AI, or related field. 5+ years of experience in applied ML, multimodal systems, or evaluation research. Strong understanding of Vision-Language Models, multimodal transformers, and embedding-based retrieval systems.

Experience designing and analyzing large-scale benchmarks and evaluation datasets. Solid background in statistical analysis, experimental design, and performance trade-off evaluation. Proficiency in Python and modern ML frameworks (e.g., PyTorch ).

Scientific Approach Ability to reason about long-tail distributions and concept coverage gaps. Experience analyzing complex multimodal system failure modes. Strong intuition for measurement of quality and evaluation bias.

Comfort operating in ambiguous, research-driven problem spaces. AI-Accelerated Research Execution Demonstrated ability to use AI coding tools and AI-assisted workflows to rapidly prototype evaluation frameworks, detection experiments, and data analysis. Ability to scale scientific insight through high-velocity experimentation.

Preferred Qualifications Experience in safety evaluation, trust & safety systems, or content moderation science. Background in retrieval-augmented systems, knowledge graph modeling, or large-scale embedding systems. Publications in venues such as NeurIPS , ICML, ICLR, CVPR, ACL, or equivalent industry research impact.

Experience translating evaluation science into deployed guardrail improvements. #FireflyGenAI About Adobe Adobe empowers everyone to create through innovative platforms and tools that unleash creativity, productivity and personalized customer experiences. Adobe’s industry-leading offerings including Adobe Acrobat Studio, Adobe Express, Adobe Firefly, Creative Cloud, Adobe Experience Platform, Adobe Experience Manager, and GenStudio enable people and businesses to turn ideas into impact, powered by AI and driven by human ingenuity.

Our 30,000+ employees worldwide are creating the future and raising the bar as we drive the next decade of growth. We’re on a mission to hire the very best and believe in creating a company culture where all employees are empowered to make an impact. At Adobe, we believe that great ideas can come from anywhere in the organization.

The next big idea could be yours. Let’s Adobe together At Adobe, we believe in creating a company culture where all employees are empowered to make an impact. Learn more about Adobe life, including our values and culture , focus on people, purpose and community , Adobe for All , comprehensive benefits programs , the stories we tell , the customers we serve, and how you can help us advance our mission of empowering everyone to create.

Adobe is proud to be an Equal Employment Opportunity employer. We do not discriminate based on gender, race or color, ethnicity or national origin, age, disability, religion, sexual orientation, gender identity or expression, veteran status, or any other protected characteristic. Learn more.

Adobe aims to make our Careers website and recruiting process accessible to any and all users. If you have a disability or special need that requires accommodation to navigate our website or complete the application process, email accommodations@adobe.com or call +1 408-536-3015. AI Use Guidelines for Interviews: Our interviews are designed to reflect your own skills and thinking.

The use of AI or recording tools during live interviews is not permitted unless explicitly invited by the interviewer or approved in advance as part of a reasonable accommodation. If these tools are used inappropriately or in a way that misrepresents your work, your application may not move forward in the process. At Adobe, we empower employees to innovate with AI — and we look for candidates eager to do the same.

As part of the hiring experience, we provide clear guidance on where AI is encouraged during the process and where it’s restricted during live interviews. See how we think about AI in the hiring experience . Expected Pay Range: Our compensation reflects the cost of labor across several U.S.

geographic markets, and we pay differently based on those defined markets. The U.S. pay range for this position is $142,700 -- $270,950 annually.

Pay within this range varies by work location and may also depend on job-related knowledge, skills, and experience. Your recruiter can share more about the specific salary range for the job location during the hiring process. In California, the pay range for this position is $187,100 - $270,950 In Washington, the pay range for this position is $168,600 - $244,200 At Adobe, for sales roles starting salaries are expressed as total target compensation (TTC = base + commission), and short-term incentives are in the form of sales commission plans.

Non-sales roles starting salaries are expressed as base salary and short-term incentives are in the form of the Annual Incentive Plan (AIP). In addition, certain roles may be eligible for long-term incentives in the form of a new hire equity award. State-Specific Notices: California : Fair Chance Ordinances Adobe will consider qualified applicants with arrest or conviction records for employment in accordance with state and local laws and “fair chance” ordinances.

Colorado: Application Window Notice If this role is open to hiring in Colorado (as listed on the job posting), the application window will remain open until at least the date and time stated above in Pacific Time, in compliance with Colorado pay transparency regulations. If this role does not have Colorado listed as a hiring location, no specific application window applies, and the posting may close at any time based on hiring needs. Massachusetts: Massachusetts Legal Notice It is unlawful in Massachusetts to require or administer a lie detector test as a condition of employment or continued employment.

An employer who violates this law shall be subject to criminal penalties and civil liability.