Interesting Data Gigs # 41: 7 Jobs, 7 Ideas

These processors remain compatible with the Arm instruction set, so there’s nothing new developers have to do to support them, but more importantly, by designing its own cores, Ampere was able to optimize these processors for its core user base of large cloud platforms. The new AmpereOne chips can support up to 192 cores, a significant upgrade from the Altra Max chips’ maximum of 128. In the AmpereOne family, every core now also gets 2MB of L2 private cache (up from 1MB) and the company added several features that ensure that every core gets access to the same amount of memory bandwidth (or not, if that’s what the user desires) and provide more fine-grained power management, for example. There are also features like process aging monitoring, which helps these vendors better manage the lifecycle of the processor and monitor it for issues, as well as advanced droop detection and security features like memory tagging and single-key memory encryption.

At Cohere, we strive to continually improve our large language models. Data acquisition is crucial for the sucess of our business. Our ability to responsibly acquire data that is accurate, relevant, and timely is key to our ability to improve the quality of our models. We strive to continuously improve our data acquisition processes and systems to ensure that we have the data we need to stay competitive and meet the needs of our customers.

We are looking for a Member of Technical Staff, Data Acquisition to join our team. This role would be part of the Data Acquisition team, which broadly provides data for training models at Cohere and is responsible for building and maintaining the infrastructure that acquires, cleans, and formats data for model training. The main responsibility of this role is to improve our internal data acquisition infrastructure, which includes data crawlers, formatters, and integrations with data providers. This role would also work closely with different teams at Cohere to support their data acquisition needs, as well as engage in more experimental work to develop highly informative data signals.