Member of Technical Staff, Research

FirstPrinciples

5d•Remote

About The Position

We are looking for a Member of Technical Staff, Research to investigate, design, test and develop state of the art (SOTA) methods and applications, which can be integrated into the broader AI engine FirstPrinciples is developing. You will collaborate with cross-functional teams and your work will flow straight into production, helping advance the way scientific research is performed. Your work will impact the wider academic community through the development of unique solutions to usher in a new era of scientific discovery. The ideal candidate has a proven track record in AI research, who can combine strategic thinking with technical depth to bring complex ideas to life.

Requirements

PhD in physics, computer science, data science, information systems, or related field.
Proven track record of conducting in-depth research on scientific AI models, symbolic models, machine learning or deep learning for scientific discovery.
Familiarity with SOTA models, best practices in model development processes, in-depth AI/ML concepts, and data infrastructure.
Comfort working closely with engineers and other technical team members.
Strong written and verbal communication skills.
Comfortable working in a startup-style, cross-functional, remote team.

Nice To Haves

Has experience with or strong interest in physics and/or fundamental science topics.
Experience conducting research on AI models in an early-stage or mission-focused environment.

Responsibilities

Research, design, and test novel, research‑specific model architectures that integrate academic literature, natural language processing (NLP), symbolic reasoning, and other methods to orchestrate the scientific process.
Prototype and build custom tokenizers for LaTeX symbols and physical units to be treated as tokens.
Explore alternatives to transformers through in-depth research and provide practical recommendations for model development.
Develop reinforcement-learning loops to enable models to run independent and internal thought experiments.
Design and automate data ingestion pipelines in collaboration with our Data Scientists & Engineers that aggregates science literature, metadata, experimental data, equations and other data sources in a robust and scalable manner.
Establish custom benchmarks to assess the models’ understanding of physical concepts, mathematical reasoning abilities, and ability to minimize hallucinations for the benefit of scientific reliability.
Refine and release datasets and baselines once internal tests are stable.
Run and track model training jobs while leading the technical team through set-up, monitoring progress, and constraining costs within budget.
Develop approaches to stage “practice runs” in a sandbox environment to develop the model’s abilities to explore ideas independently while logging results for later review.
Develop a framework to evaluate the models’ learning using visual and statistical tools to spot patterns and blind spots.
Add guard-rails and tests that flag poor quality model output.
Maintain internal tools to track lists of known issues, noting failures, clear fixes, and improvements to be integrated into future development.
Work with the engineering team to ensure product feasibility and robust architecture.
Translate technical trade-offs to non-technical stakeholders in clear terms.
Present findings in clear updates to the technical team in order to keep the broader team appraised of progress against research milestones.