Member of Technical Staff (ML)
Solstice · New York City · $200,000–$350,000 (stated in the posting)
Why you should join us
Solstice builds AI software that helps life sciences teams create, refine, and review pharmaceutical marketing content. We have worked with leading life sciences manufacturers and more than 50 pharma brands. Our backers include Transformation Capital, Twelve Below, Virtue, and the founders of Datavant, Commure, and Paradigm.
The technical problems are concrete: find the right evidence in complex scientific material, identify claims accurately, and improve generated content using reviewer feedback and real product outcomes. You will work with proprietary data and a team that can put your improvements in front of customers.
About the role
As a Member of Technical Staff in Machine Learning, you will improve the models, representations, and data that power Solstice's AI agents for creating, refining, and reviewing pharmaceutical marketing content. Your work will help agents find supporting evidence, use tools accurately, and incorporate reviewer feedback into better outputs.
You will investigate failures across complete agent runs, form a testable hypothesis, implement a change, and establish whether it improves the customer's task. The intervention might be better training examples, a retrieval or ranking improvement, a different model or context strategy, or targeted fine-tuning. You will choose an approach based on evidence and test its quality, reliability, latency, and cost before taking it into production.
You will work directly with Applied AI engineers, who own the complete feature and workflow, and Fullstack AI engineers, who build the shared runtime, APIs, and product experiences. You will build prototypes and production components together, with primary responsibility for model and data improvements and the evidence that they generalize.
What you'll be working on
Improve model behavior inside agents. Diagnose failures in evidence selection, tool choice and arguments, and revisions based on feedback. Work with Applied AI to separate model or data limitations from missing context, tool failures, or workflow design problems.
Make retrieval and grounding more accurate. Develop and compare retrieval, reranking, and claim-to-evidence matching approaches across scientific documents and approved content. Measure whether the evidence actually supports the claim and helps the agent complete the task.
Turn agent runs and reviewer feedback into useful data. Build curated datasets from successful and failed runs, corrections, and targeted examples. Track provenance, resolve ambiguous labels, and keep held-out evaluation cases separate from training and tuning.
Implement and test substantive improvements. Compare model choices, representations, context strategies, and learning methods with credible baselines and controlled experiments. Use targeted fine-tuning or other model adaptation when the available data and results justify it.
Measure complete-task outcomes. Evaluate intermediate tool behavior and the resulting artifact. Check repeated-run reliability and generalization to unfamiliar documents and task variants. Calibrate automated judgments against human review and track quality alongside latency and cost.
Ship improvements with the product team. Integrate changes into production Python services and agent workflows. Use versioned data and models, staged releases, quality monitoring and rollback criteria, then investigate whether offline gains hold up in use.
What we're looking for
Strong ML fundamentals and hands-on implementation skills. You can reason about learning objectives, optimization, regularization, data quality, and generalization.
Strong Python and hands-on experience building or improving LLM systems, retrieval systems, or agents. You can inspect tool-call traces, reproduce failures, and implement the experiment or production change needed to investigate them.
Evidence of taking a data or model improvement into production and understanding what happened after deployment.
Sound experimental judgment: explicit hypotheses, credible comparisons, meaningful baselines, and care with leakage, noisy labels, and misleading aggregate metrics.
Depth in at least one relevant area, such as retrieval and ranking, representation learning, model adaptation, or learning from human feedback, with the ability to apply that depth to multi-step agent tasks.
Judgment about the source of a failure. You can explain when to change the data or model, when to improve context or a tool interface, and how to design a comparison that distinguishes those explanations.
Practical software engineering skills for reproducible experiments, tested components, production integration, and debugging.
Clear communication and personal ownership. You can explain the tradeoffs, work with product and engineering partners, and follow through when results challenge your assumptions.
Experience with scientific documents, human review feedback, or regulated enterprise products is useful. Healthcare or pharma experience is welcome but not required. We value demonstrated implementation and experimental judgment; a particular degree, publication record, or framework is not a substitute for that evidence.
Benefits
Health, dental, and vision insurance
Ground-floor equity opportunity
Visa sponsorship for O-1, H-1B, and TN visas
Unlimited PTO
401(k) with a 4% company match
In-office lunches five days a week
Professional growth stipend
Relocation support
Location and compensation
New York City. Full time. On-site.
The base salary range is $200,000 to $350,000, plus equity, depending on experience and scope of responsibility.