data refining for better AI
Explore your training datasets
Backed by cognitive sciences, Bunka interface relies on topographical-like maps to explore textual datasets in a new and intuitive way
Refine your data
By combining Visual Topic Modelling with advanced cluster analysis, Bunka improves data quality, helps remove outliers and redundant data
Improve model quality
With greater data quality comes increased model performance, reduced fine-tuning time and regulation compliance
.png)
.png)
Bunka simplifies dataset exploration
01
State-Of-The-Art Visualizations
Bunka provide an intuitive and innovative way of exploring textual datasets, using cognitive-sciences backed maps and advanced algorithms to provide an overview of your data.
03
Advanced Metrics
Bunka improves traditional visualizations by adding another fully customizable dimensional layer. Whether ML engineers want to analyze the perplexity, the toxicity or several other text quality metrics of their training data, Bunka got them covered.
02
Modular Technology
Bunka technolgy is fully adaptive and connects to various frameworks, whether it is the last LLM model, a new SOTA embedder or another clustering algorithm.
04
Connected with AI Community
At Bunka, we keep experimenting with new ways to understand textual data quality. We ensure that our technology is up-to-date by partaking in many Open Source projects and contributing to build a better data quality ecosystem.
Meet the team
Charles de Dampierre
co-founder, CEO
PhD in computational social sciences and Master in cognitive sciences @ENS, HEC Paris Alumni
Strategy major @HEC, Linguistics and cognitive sciences major @ENS, entrepreneur
Louis-Marie Lorin
co-founder, COO
Research director @CNRS, Thesis director @ENS, cognitive science expert, +4500 citations
Nicolas Baumard
co-founder, CSO
Cognitive sciences @ENS, specialized in AI and Philosophy
Ciprian Bangu
AI Intern
Elias Showk
Lead Tech
Full stack developer, UTC Alumni, former senior developer @HuggingFace
SUPPORTED BY
.png){kind=logo}
.png)