About Anosys

Anosys is an AI operational intelligence platform that connects observability, agent traces, evaluations, costs, user behavior, and business outcomes. We help teams understand how their systems perform, investigate complex issues, and identify opportunities to improve reliability, efficiency, and product experience.

We’re expanding our applied science team to advance the scientific methods and applied AI capabilities behind the platform. This is an opportunity to work on consequential technical problems, influence product decisions, and see your work used in production.

About the Role

As a Senior Applied Scientist at Anosys, you will advance how we learn from complex operational and behavioral data. Your work will span areas such as causal modeling, opportunity identification, LLM adaptation, evaluation, and statistical learning.

You will translate ambiguous problems into clear research questions, develop and evaluate approaches, and collaborate with engineering and product teams to deliver reliable improvements. You’ll balance scientific rigor with practical constraints—including data quality, latency, cost, interpretability, and customer usefulness.

We’re looking for depth in one or more relevant disciplines, strong experimental judgment, and the ability to connect research to real-world outcomes.

What You’ll Do

• Advance causal modeling and inference. Develop and evaluate methods for understanding relationships between system behavior, user activity, and business outcomes. Make assumptions explicit, assess uncertainty, and distinguish correlation from defensible causal conclusions.
• Identify meaningful opportunities. Apply statistical and machine-learning methods to uncover patterns, inefficiencies, behavioral differences, and opportunities for improvement. Assess findings for significance, potential impact, and practical relevance.
• Improve LLM capabilities. Research and apply model adaptation, fine-tuning, and optimization techniques for analytical and domain-specific tasks. Evaluate improvements against credible baselines, considering accuracy, robustness, latency, and cost.
• Strengthen evaluation methodology. Design and improve evaluations for models, agents, and AI-powered workflows. Develop representative datasets, meaningful metrics, and reproducible experiments that expose limitations and measure progress.
• Advance observability and product analytics. Apply scientific methods to problems involving anomaly detection, distribution shifts, behavioral analysis, segmentation, attribution, and performance diagnosis across heterogeneous data.
• Turn research into production improvements. Work closely with engineers to translate successful approaches into maintainable, scalable capabilities. Assess performance under real-world conditions and use feedback to guide further improvements.
• Shape research priorities. Partner with product and customer-facing teams to understand important problems, evaluate competing approaches, and focus effort where scientific advances can create meaningful value.
• Communicate with clarity. Document methods, assumptions, experiments, and conclusions. Explain tradeoffs to technical and nontechnical colleagues, and contribute to a culture of rigorous, constructive research.

What We’re Looking For

• Strong foundations in machine learning, statistics, experimental design, and quantitative reasoning.
• Deep expertise in one or more relevant areas, such as causal inference, LLM adaptation, evaluation science, statistical learning, anomaly detection, time-series analysis, or product analytics.
• Experience conducting rigorous research or applied experimentation, with evidence of translating findings into working systems or meaningful practical outcomes.
• Strong Python skills and proficiency in SQL, with experience using relevant machine-learning frameworks and data-processing tools.
• The ability to work with noisy, incomplete, high-dimensional, or changing data—and recognize when the available evidence does not support a conclusion.
• Sound evaluation practices, including appropriate baselines, reproducibility, uncertainty assessment, and controls against data leakage and overfitting.
• Strong independent judgment: you can frame an ambiguous problem, prioritize the important questions, and make progress without requiring a fully specified solution.
• Clear written and verbal communication, with the ability to collaborate effectively across research, engineering, and product.

An advanced degree in Computer Science, Statistics, Applied Mathematics, Machine Learning, or a related field is valuable, as is equivalent research or industry experience. Publications, shipped systems, open-source contributions, and substantial technical projects are all relevant evidence of your capabilities.

Experience That Would Be Especially Valuable

• Causal models, experimental and observational study design, or treatment-effect estimation.
• Fine-tuning, adapting, or evaluating LLMs and agentic systems.
• Product analytics, user behavior modeling, segmentation, experimentation, or business process analysis.
• Observability, distributed tracing, telemetry, or production ML monitoring.
• Learning from limited labels, expert feedback, or heterogeneous data sources.
• Experience improving models under production constraints, including efficiency, reliability, and data-access requirements.
• A record of working effectively in a growing company where research priorities respond to product and customer needs.

How We Work

We value curiosity, intellectual honesty, and ownership. We expect scientists to challenge assumptions, investigate unexpected results, and choose methods based on evidence rather than novelty.

You’ll have meaningful responsibility for your work and close collaboration with the people who bring it into production. Strong ideas matter; so do clear reasoning, reliable execution, and the ability to improve the experience of the teams using Anosys.

We support sharing research through technical writing, open-source contributions, and publications where appropriate.

What Success Looks Like

Success means advancing the platform through research that is well-founded, reproducible, and useful. You’ll be evaluated on the quality of your judgment, the reliability of your methods, your contribution to production improvements, and how effectively you help the team solve important problems.