Research Scientist / Research Engineer

Adaption's AutoScientist automates model fine-tuning with closed-loop training outperforming human-designed configurations. Published: May 14, 2026 at 8:18 am Updated: May 14, 2026 at 8:18 am Edited and fact-checked: May 14, 2026 at 8:18 am Adaption unveils AutoScientist, a system that automatically customises AI models by optimising both training data and learning processes for specific tasks. Adaption, an AI startup founded by former Cohere Vice President of Research Sara Hooker, has introduced a new system called AutoScientist, designed to automate the process of tailoring AI models to specific tasks by jointly optimising both training data and learning configurations. The system is positioned as a step toward automating AI research and development workflows, with the aim of reducing the manual effort typically required in model fine-tuning and experimentation. AutoScientist is described as an end-to-end framework that co-optimises datasets and training recipes simultaneously, iterating through a closed loop in which both data selection and model training parameters are continuously adjusted. The process is intended to continue until performance stabilises around a defined objective, effectively allowing the system to refine both what the model learns from and how it learns it without constant human intervention. According to the company, the tool is intended to reduce the time required to move from an initial concept to a deployed, customised model, potentially compressing development cycles from weeks to hours. It is also presented as a mechanism that broadens access to model customisation beyond machine learning specialists, enabling users without deep technical expertise to influence not only prompts but also the underlying behaviour of trained systems. The approach is framed as particularly relevant for organisations seeking to fine-tune models for domain-specific language, structured outputs, or efficiency constraints such as latency and cost, while leveraging proprietary datasets more effectively within AI systems. Internal evaluations referenced by the company suggest that AutoScientist demonstrates improved performance compared with baseline models across a range of dataset sizes between 5,000 and 100,000 examples, as well as across multiple model architectures available for fine-tuning. Reported results indicate consistent gains regardless of domain, with performance measured using in-house evaluations tailored to specific vertical applications. Further comparisons presented in the evaluation framework indicate that AutoScientist achieved higher average performance than configurations designed by human researchers, including experienced AI engineering staff. In these tests, human experts selected training setups based on their knowledge of model architecture, dataset characteristics, and domain requirements, while AutoScientist was given the same inputs along with the ability to iteratively refine its own configurations using historical run data. Under these conditions, aggregate outcomes reportedly improved from 48 percent to 64 percent when using the automated system, with an average performance uplift of approximately 35 percent across experiments. AutoScientist shows cross-domain stability while aiming to democratise frontier model fine-tuning. Additional benchmarking across multiple application areas suggests that the system is not strongly sensitive to specific domains, with gains observed across eight different verticals. The company reports that this consistency is notable given that many traditional fine-tuning approaches tend to underperform outside narrow or highly curated settings, whereas AutoScientist reportedly delivers more stable improvements across varied tasks and datasets. The system is positioned as part of a broader effort to automate model development processes, particularly in areas involving long-horizon reasoning, which remains a persistent challenge in AI reliability. The developers indicate that AutoScientist represents an early step toward reducing the need for manual intervention in model training pipelines, with future research directions focused on enabling more immediate forms of adaptation that may not require traditional training cycles. Alongside its technical objectives, the release is also framed as an effort to broaden access to model customisation, allowing a wider range of users to shape AI systems for specific applications. The tool is being made available free of charge for an initial 30-day period. The broader aim, according to the framing provided, is to reduce barriers to AI model development and expand the ability to create tailored systems beyond a small group of specialised researchers concentrated in major laboratories. A key contextual argument highlighted in the announcement is that only a small number of people globally possess the expertise required to properly train and fine-tune frontier AI models, with most of this knowledge concentrated within a limited number of major research laboratories. It is suggested that if a system such as AutoScientist is able to successfully automate aspects of this expertise, the process of building customised models for individual organisations and specific use cases could become more accessible and practically achievable. Disclaimer. In line with the Trust Project guidelines, please note that the information provided on this page is not intended to be and should not be interpreted as legal, tax, investment, financial, or any other form of advice. It is important to only invest what you can afford to lose and to seek independent financial advice if you have any doubts. For further information, Mpost Media Group suggest referring to the terms and conditions as well as the help and support pages provided by the issuer or advertiser. MetaversePost is committed to accurate, unbiased reporting, but market conditions are subject to change without notice. Alisa, a dedicated journalist at the MPost, specializes in crypto, AI, investments, and the expansive realm of Web3. With a keen eye for emerging trends and technologies, she delivers comprehensive coverage to inform and engage readers in the ever-evolving landscape of digital finance. Alisa Davidson Hot Stories by Alisa Davidson May 15, 2026 by Alisa Davidson May 15, 2026 by Alisa Davidson May 15, 2026 by Alisa Davidson May 15, 2026 by Alisa Davidson May 15, 2026 by Alisa Davidson May 15, 2026 by Alisa Davidson May 15, 2026 by Alisa Davidson May 15, 2026

Adaption. Expand Your World Apr 14, 2026 Today, Adaption Labs, Inc. is announcing a new feature in Adaptive Data that is called: Expand Your World. Most datasets are built for a narrow slice of the world. Not due to intent, because building data is hard, slow, and expensive. Teams make pragmatic choices. They start with the languages they know, the markets they're already in, the users already at their door. What gets left out tends to stay left out. The result is AI that works well for some and barely at all for others. Not because the problems are harder in those communities. The data never showed up. The World Doesn't Speak One Language. Your Data Shouldn't Either. A dataset built on ten languages may feel comprehensive, but the world speaks far more than ten. Adaptive Data supports 242 languages and localizations, giving your dataset a reach that most teams couldn't build on their own. The problem isn't just coverage, it's depth. Within a single language, regional dialects and cultural variation shape meaning in ways that matter. A model trained on one variant of a language will often fail users who speak another. Language diversity in your dataset isn't a nice-to-have, it's foundational. The data you train on defines the boundaries of what your model can understand, represent, and get right. If that data skews toward a handful of languages, your model inherits those blind spots. No amount of fine-tuning later fully closes the gap. Getting language coverage right at the data stage is the only way to build models that are capable across the communities they serve. The standard solution to this has been to hire more annotators and build more pipelines. Brute force doesn't scale to the whole world. Adaptive Data does. The Fastest Way to Global Coverage Expand Your World takes what you already have and multiplies its reach. Starting from as few as 10 examples in a single language, Adaptive Data generates up to 2,420 diverse, high-quality examples across all 242 languages and localizations. The workflow requires no extra pipelines, no annotation overhead, and no additional lift from your team. This is not a late-stage consideration or a bolt-on step. It is a fundamental expansion of what your dataset can cover, built into the data layer from the start. Expand Your World is available to all Adaptive Data users today. Its Research Grant Program provides platform access for teams exploring Adaptive Data systems and global language coverage. Priority is given to applications focused on advancing open science or positively shaping the public good. Early-stage teams can also explore Adaption for Startups, a program designed to help startups build with Adaptive Data from the ground up.

Adaption. Blueprint: A Specification Layer for Adaptive Data Mar 17, 2026 Two weeks ago Adaption Labs, Inc. announced Adaptive Data, a platform to allow everyday builders to control and adapt their data. Adaptive Data is a fundamental shift: Adaption Labs, Inc. treat the data space as dynamic and fully malleable. Today Adaption Labs, Inc. is adding a new powerful capability to Adaptive Data for builders everywhere to shape AI. Adaption Labs, Inc. call its release Blueprint. Blueprint allows you to steer the data space towards any goal you want. Think of it as the preferences sheet you give to a concierge service before you trust AI to come back with a tailored itinerary. Blueprint efficiently steers data towards your specific desirable properties, and automatically learns penalties if AI violates any of your rules. The field has made extraordinary progress on what models can do. The harder, quieter question - how to make AI behavior durable, transferable, and precisely defined - has received far less attention. Adaption Labs, Inc. think that's the next important frontier. Blueprint is its first move toward it. Why Specification Is an Unsolved Problem Telling an AI system what to do is easy. Ensuring it reliably does that, and only that, across contexts, inputs, and time is not. The dominant approach is prompt engineering: iterative, informal, and local. It works until it doesn't. Requirements accumulate as workarounds. Constraints that held in one context quietly fail in another. There is no principled mechanism for encoding what good behavior looks like in a way that is explicit, auditable, and persistent. This is a specification problem. And unlike capability, it doesn't get easier as models get more powerful. If anything, the gap between what a system can do and what it should do in any given context widens. A New Layer in Adaptive Data Adaption Labs, Inc. built Blueprint because Adaption Labs, Inc. firmly believe data that evolves with the world is only useful if it evolves the right way. With this release, Adaption Labs, Inc. is bringing innovations in data optimization typically reserved for frontier labs to everyone. Blueprint lets you define your goals - length, tone, safety thresholds, custom content policies - and have those automatically changed to objectives which optimize every dataset the platform produces. When your requirements change, your data automatically evolves with them. The design principle is simple but consequential: data configuration should be foundational, not cosmetic. Your requirements become the specification every data point is evaluated against. Not a preference layered on top. A structural constraint built in from the start. Adaption Labs, Inc. think this reframing matters. Treating data behavior as a first-class property - something you define, version, and enforce - is a different posture toward data systems than the field has generally taken. The Way Forward The field has made remarkable progress on what AI systems can do. The next frontier is making them controllable - not just capable. That requires innovation at every layer: in how data is shaped and specified, in how intelligence generalizes across contexts, and in how humans interact with systems that are no longer static. Blueprint is one piece of that. But its underlying conviction is larger genuine control over AI behavior is both an unsolved research problem and one of the most important ones to work on. At Adaption, Adaption Labs, Inc. started this journey because Adaption Labs, Inc. believe today's AI is backwards. Most systems are static, expensive, and slow to change. Exceptional use cases, the ones that matter most in real world contexts are treated like outliers. Adaption Labs, Inc. is setting out to change that. Intelligence should not arrive preconfigured. AI should evolve and learn continuously. That starts with data. Everything intelligent adapts. So should AI. Blueprint is available to all Adaptive Data users around the world today. Its commitment is to empower everyday individuals, developers and enterprises to shape and control their AI. Its research grant program provides access to the platform for teams exploring adaptive data systems, behavioral specification, and AI control. Mar 17, 2026