June 16, 2025
By Anjali Kocchar
In an industry where artificial intelligence is evolving at breakneck speed, Together AI has unveiled its latest innovation: an open-source AI agent designed to function as a full-fledged data scientist. With this release, developers, researchers, and enterprises gain access to a powerful new tool one that can reason about data, execute code, generate insights, and iterate on results entirely driven by open-source models and transparent workflows.
Built on top of the powerful Together AI stack, the data-scientist agent integrates code execution and reasoning through a ReAct-style iterative process. The approach mirrors how a skilled human analyst works: the agent proposes a hypothesis, writes code to extract or visualise data, assesses the results, self-corrects, and refines its strategy. The result is a more reliable and autonomous agent capable of producing end-to-end data analysis streamlining tasks that once demanded deep expertise and hours of manual work.
This launch reflects Together AI’s broader mission: to democratise AI by empowering users to build, fine-tune, and deploy complex workflows with open-source models in secure, serverless environments. The company already supports advanced reasoning models like DeepSeek R1, offers privacy-first infrastructure, and backs its tools with a research agenda that includes Mixture of Agents and other innovations.
At a practical level, the new agent includes a code interpreter interface that connects to datasets, applies exploratory data analysis techniques like summary statistics and visualisations, and packages findings into structured outputs such as charts and tables. Usage is straightforward: developers call the Together API with their data and prompt, and the agent handles both code execution and iterative improvement until achieving meaningful insight .
Behind the scenes, the agent leverages Together AI’s infrastructure: serverless inference with per-token pricing, secure data controls for privacy compliance, and scalable GPU clusters optimised with FlashAttention and other performance enhancements on NVIDIA hardware.
This is no mere tech demo. The agent makes advanced analytics accessible to non-specialists data-savvy business teams, domain experts, and even students can now deploy a data-science-grade assistant without writing any code. At the same time, seasoned data scientists can customise prompts and workflows to meet professional standards.
It signals a future where AI-powered analytics are transparent, collaborative, and equitable made possible by combining powerful infrastructure with open-source intelligence.
About the author
Anjali Kochhar covers cryptocurrency and blockchain stories in India as well as globally. Having been in the field of media and journalism for over four years now, she has developed a sharp news sense and works hard to present information that goes beyond the obvious. She is an avid reader and loves writing on a wide range of subjects.
