ML Applied Researcher (Foundation Models)

Akkar; San Francisco (CA); Hybrid; USD 180,000 - 280,000; Full time

Job summary

Akkar is seeking a Senior ML Research Scientist to advance general-purpose foundation models, collaborating with research leaders to learn across data types and translate results into product features.

You will own research problems from hypothesis to training, design experiments, and work on multimodal data at scale in a hybrid Bay Area setup.

Qualifications

• Hands-on foundation-model training experience at substantial scale, including training runs involving billions of tokens or comparable large-scale multimodal data.
• A clear understanding of training stages, data choices, model design and evaluation, with evidence of the decisions you personally owned.
• Ability to develop original hypotheses, run rigorous experiments and independently decide what to investigate next.
• Strong practical implementation skills, with evidence of taking research beyond a prototype towards usable systems.
• Ability to explain complex technical work clearly in a live discussion, including why an approach worked or failed.
• Availability to work with the team in the Bay Area on a hybrid basis.

Responsibilities

• Own research problems from the initial hypothesis through experiment design, model training, evaluation and iteration.
• Develop and test improvements to model architectures, training methods and data strategies.
• Train foundation models at meaningful scale and investigate their capabilities, limitations and failure modes.
• Explore new modelling approaches and ways to extend models across different data types.
• Work with research and ML systems colleagues to turn promising results into implemented, tested capabilities.
• Explain your reasoning, findings and trade-offs clearly, helping shape the team’s next research decisions.

Skills

Foundation model training

Experiment design

Model evaluation

Multimodal learning

Communication skills

Hybrid collaboration

Education

PhD in machine learning

Job description

Work Environment: Hybrid (2–3 office days per week)

The Company and Opportunity

Akkar is partnering with a well-funded AI startup founded by former Google researchers and technology leaders, with a track record of taking new technology from research into widely used products.

The company has already secured investment and delivered early enterprise deployments. As it expands its research and commercial activity, this is an opportunity to join a small team with substantial room to influence the technology and grow your responsibilities.

You will work closely with experienced research leaders on general-purpose foundation models, exploring how they can learn across different types of data and support real-world applications. There are open questions to investigate, scope to propose your own approaches, and a clear route for successful research to become part of the product.

What you will own

• Own research problems from the initial hypothesis through experiment design, model training, evaluation and iteration.
• Develop and test improvements to model architectures, training methods and data strategies.
• Train foundation models at meaningful scale and investigate their capabilities, limitations and failure modes.
• Explore new modelling approaches and ways to extend models across different data types.
• Work with research and ML systems colleagues to turn promising results into implemented, tested capabilities.
• Explain your reasoning, findings and trade-offs clearly, helping shape the team’s next research decisions.

What you will bring

• Hands‑on foundation‑model training experience at substantial scale, including training runs involving billions of tokens or comparable large‑scale multimodal data.
• A clear understanding of training stages, data choices, model design and evaluation, with evidence of the decisions you personally owned.
• The ability to develop original hypotheses, run rigorous experiments and independently decide what to investigate next.
• Strong practical implementation skills, with evidence of taking research beyond a prototype towards usable systems.
• The ability to explain complex technical work clearly in a live discussion, including why an approach worked or failed.
• Availability to work with the team in the Bay Area on a hybrid basis.

Your experience should show depth in training and developing foundation models; small‑dataset fine‑tuning alone will not meet the technical bar.

Useful additional experience

• A PhD in machine learning or a related field, or equivalent industry research achievements.
• Strong contributions to publications at leading ML conferences.
• Multimodal learning, representation learning or self‑supervised learning.
• Time‑series, sensor data, robotics, autonomy or other real‑world ML applications.

Strong language‑model researchers are welcome. Recent PhD graduates will also be considered where publications and substantial research internships demonstrate model‑training depth and practical ability.

How you like to work

You are articulate, curious and comfortable tackling problems without a ready‑made answer. You have a point of view and are happy to challenge an idea constructively, while listening carefully and working collaboratively with others.

You take ownership of your work, make progress independently and know when to bring colleagues into a decision. A small startup team, changing priorities and a mix of creative research and practical delivery appeal to you.