Where this role sits

This is a research role focused on the models at the foundation of collective intelligence. You study, adapt, and advance the foundational models that power how Sapience AI understands language, knowledge, and reasoning.

You work where research meets the platform: designing experiments, evaluating models, adapting them to the demands of professional communities, and feeding what you learn into the COGENT architecture and MINERVA.

You bring scientific rigor to a fast-moving field, and you turn that rigor into advances the product can actually use.
Why this role exists

The quality of collective intelligence depends on the models beneath it. How well the platform understands a community's language, grounds its answers, and reasons over knowledge starts with foundational model work done well.

The field moves quickly, and not every advance is real or ready. Someone has to separate genuine progress from noise and turn the real advances into something the platform can rely on.

The Foundational Model Research Data Scientist does that. You run the experiments, evaluate honestly, and translate frontier progress into dependable capability for Sapience AI.
What you will own (Areas of Responsibility)

You hold seven areas of responsibility across foundational model research. Each one is yours to set direction on, build, and measure.
1. Foundational model research and experimentation

• Design and run experiments on foundational models relevant to collective intelligence.
• Investigate how models understand language, ground answers, and reason over knowledge.
• Turn open questions into experiments with clear hypotheses and honest results.

2. Model adaptation and fine-tuning

• Adapt foundational models to the language and needs of professional communities, including fine-tuning and alignment where it helps.
• Improve grounding and reduce confident errors in domain settings.
• Balance capability against cost, latency, and the constraints of production.

3. Evaluation and measurement

• Build rigorous evaluation for what matters here: accuracy, groundedness, safety, and trust.
• Design evaluations that reflect real community needs, not just public benchmarks.
• Keep the organization honest about what a model can and cannot do.

4. Data for models

• Partner with data engineering on the datasets that training and evaluation depend on.
• Handle data thoughtfully, including quality, bias, and protection of sensitive community knowledge.
• Build the evidence base that makes model claims defensible.

5. Integration with COGENT

• Feed model advances into the neuro-symbolic COGENT architecture, and study how neural and symbolic methods work together.
• Help decide where a foundational model belongs and where structure should carry the load.
• Turn research into behavior the platform can rely on.

6. Staying at the frontier

• Track the fast-moving foundational model field and separate real progress from hype.
• Bring in advances that matter and set aside those that do not.
• Share knowledge so the whole organization stays current.

7. Responsible and trustworthy AI

• Study and reduce the failure modes that erode trust, including hallucination and bias.
• Build toward models whose answers members can trust and trace.
• Treat safety and trust as part of the research, not a later concern.

AI-augmented ways of working

AI is both your subject and your tool. You use AI to accelerate literature review, code experiments, and analysis, while holding the scientific rigor that makes results trustworthy.

The standard is human in partnership: AI accelerates the work, you own the judgment, the interpretation, and the call. The people who create the most value here are not the ones producing the most output. They are the ones turning evidence into clear, durable decisions.
What this role is not

To keep the boundary clear:

• This is not a pure publications role. Your research is measured by advances the platform can use, not papers alone.
• This is not an ML infrastructure role. You partner with infrastructure on training and serving, but your focus is the models and the science.
• This is not a data engineering role. You partner with data engineering on datasets; you do not own the data platform.
• This is not a benchmark-only role. You are measured on trustworthy capability in real community settings, not leaderboard scores.

What success looks like

We measure this role on outcomes the team can see:

• Real advances. Your work improves how the platform understands, grounds, and reasons, in ways members feel.
• Honest evaluation. The organization has a clear, trustworthy picture of what the models can do.
• Better grounding. Confident errors go down, and answers become more traceable.
• Frontier awareness. Sapience AI adopts the advances that matter and skips the ones that do not.
• Research into product. Your findings become dependable behavior in COGENT and MINERVA.
• Trust by design. Safety and trust improve as a result of your research, not despite it.

Who you are
Required qualifications

• A strong research background in machine learning, NLP, or a related field, with a graduate degree or equivalent experience.
• Hands-on experience with foundational models and modern LLMs, including training, fine-tuning, or evaluation.
• Rigor in experiment design, evaluation, and honest interpretation of results.
• Strong Python and modern ML frameworks.
• The ability to turn research into advances a product can use.
• Care for safety, bias, and trust in model behavior.
• Clear written communication of technical findings.

Preferred qualifications

• Publications, patents, or shipped systems in foundational models or applied NLP.
• Experience with retrieval-augmented generation and grounding.
• Familiarity with neuro-symbolic methods and knowledge graphs.
• Experience adapting models to specialized domains.
• Experience handling sensitive or regulated data responsibly.

How you work

• You start from a clear question and name it before reaching for a method.
• You are honest about results, including negative ones.
• You balance frontier ambition with what production can bear.
• You treat trust, safety, and bias as part of the science.
• You share knowledge and lift the people around you.

Skills & Competencies

• Foundational model research, fine-tuning, and alignment.
• Evaluation design for accuracy, groundedness, and safety.
• Experiment design and rigorous analysis.
• Grounding and retrieval-augmented methods.
• Working with sensitive data responsibly.
• Translating research into product-ready advances.
• Clear technical writing and communication.

Services & Tools Experience

• PyTorch or equivalent deep-learning frameworks.
• LLM training, fine-tuning, and serving tooling.
• Experiment tracking and evaluation frameworks.
• Retrieval, embeddings, and vector systems.
• Distributed training and cloud or GPU environments.
• Python as the primary language, plus data and analysis tooling.
• Integration with the COGENT architecture and the MINERVA platform (trained on the job).

Prior Experience & Background

• Prior research or applied science work on foundational models, LLMs, or NLP.
• A track record of experiments that led to real advances or sound decisions.
• Experience bridging research and engineering.
• Industry research experience in a fast-moving AI setting is a plus.

Cross-functional partners

You work most closely with Neuro-Symbolic AI, Applied AI, ML Infrastructure, and Data Engineering. You feed foundational model advances into the COGENT architecture and the MINERVA platform.
How we hire

We review every application, and we encourage you to apply even if you do not match every line above. Research shows that talented people, especially those from underrepresented communities, often hold back when they do not meet every qualification. If that is the only thing holding you back, apply anyway.

Compensation

Base Salary: $204,000 - $216,000 + early stage equity

Generous health and wellness benefits