Austin, Texas, Feb. 10, 2026 (GLOBE NEWSWIRE) — Sophelio, an award-winning applied AI and machine-learning company, has just announced the launch of the Data Fusion Labeler (dFL) , a platform designed to harmonize, label, and prepare complex multimodal time-series data for machine learning and advanced analytics — reducing weeks or months of manual data preparation into hours. By accelerating data readiness, dFL enables teams to move models into validation and production faster, shortening overall deployment timelines.
Sophelio Introduces the Data Fusion Labeler (dFL) for Multimodal Time-Series Data – The only labeling and harmonization studio built for multimodal time-series with full provenance you can replay
“dFL reflects both where we came from and where we’re going,” said Craig Michoski, Co-Founder of Sophelio. “It grew out of real-world fusion research, where reproducibility and data integrity are essential. We’ve evolved it into a general-purpose platform that helps teams turn raw, fragmented signals into reliable datasets in minutes instead of weeks or months.”
The launch follows Sophelio’s recent name transition and marks a major milestone in the company’s evolution. dFL has been in development for more than a year, created to meet the extreme data requirements of fusion energy research— an environment where data quality, synchronization, and reproducibility are critical. Since then, the platform has been expanded and refined to support a broad range of data-intensive applications.
Following its earlier pre-announcement, the Data Fusion Labeler is now available, with a beta version already live and in use today. Early users have access to dFL’s core capabilities for harmonizing, aligning, labeling, fusing, and exporting complex multimodal time-series data with full provenance.
Today, dFL enables teams across advanced manufacturing, energy systems, robotics, climate science, and applied research to transform heterogeneous, noisy, and asynchronous sensor data into coherent, ML-ready datasets.
As interest in data harmonization for multimodal time-series data grows, teams are increasingly evaluating a wide range of data labeling and preparation tools. To help frame this landscape, Sophelio recently published a neutral overview of leading approaches and platforms used across the field, highlighting the challenges of applying traditional labeling tools to real-world sensor data.
From Raw Signals to Reproducible Datasets
Built for teams working with complex signals (either lightweight or at scale), dFL unifies data ingestion, preprocessing, visualization, automated and manual labeling of complex time-series and sensor data, and machine-learning-ready export into a single, reproducible workflow.
Every transformation and every label is captured with deterministic, end-to-end provenance, enabling exact replay, auditability, and confidence in downstream models—whether teams are iterating locally or collaborating across organizations.
A Signal-First Approach to Data Preparation
Unlike traditional notebook-based workflows or visualization tools, dFL recognizes that how you prepare data—and the order in which you prepare data matters, since common preparation steps can produce different results if applied in a different sequence.
dFL enforces and records the exact sequence of operations, preserving the semantic meaning of the data and ensuring results can be reproduced consistently across datasets, teams, and time.
Key Capabilities of the dFL
- Multimodal data ingest for heterogeneous sensor, simulation, and log data
- Built-in harmonization engine with configurable trimming, gap filling, resampling, smoothing, and normalization
- Manual and automated labeling via interactive tools, statistical detectors, physics-informed methods, and ML-based autolabelers
- Advanced visualization, including time-series, spectral, distribution, and correlation views
- Python SDK and extensibility for custom graphs, preprocessing steps, and autolabeling logic
- Provenance-rich export to CSV, Parquet, and ML pipelines with complete metadata
- Integrated prompt engineering interface for active and agentic development projects
In production deployments, dFL has demonstrated order-of-magnitude reductions in time-to-analysis. Across multiple real-world deployments, users have reported that dFL enables hundreds of records per hour to be consistently labeled, compared to only a handful per day using prior workflows. Additional technical details are available in Sophelio’s recent preprint .
Research Foundations
The design and impact of the Data Fusion Labeler are detailed in the technical paper “The Data Fusion Labeler (dFL): Challenges and Solutions to Data Harmonization, Labeling, and Provenance in Fusion Energy,” now available on arXiv.
The paper presents dFL as a unified, operator-order-aware workflow for uncertainty-aware data harmonization and provenance-rich labeling at scale, demonstrating greater than 50× reductions in time-to-analysis using real fusion energy data. Read the paper: https://arxiv.org/abs/2511.09725
Availability
The Data Fusion Labeler (dFL) is available today, with plans ranging from a free discovery tier to enterprise deployments offering advanced data fusion capabilities, expanded API access, and premium support.
Early access to dFL beta is available at: https://dfl.sophelio.io
About Sophelio
Sophelio is an applied AI and machine-learning company focused on transforming complex, high-stakes sensor data into trustworthy, ML-ready datasets. Originating in fusion energy research, the company brings deep expertise in signal-first analytics and processing, data harmonization, and reproducible workflows to industries including advanced manufacturing, robotics, energy, and scientific research. Sophelio specializes in building products and solutions for complex, high-stakes time-series and multimodal data, where accuracy, reproducibility, and traceability are critical. Formerly Sapientai, the company develops tools for data harmonization, labeling, fusion, and provenance that support machine learning, scientific discovery, and operational decision-making across fusion energy, advanced manufacturing, robotics, and other data-intensive domains. Sophelio’s flagship software, the Data Fusion Labeler (dFL), enables teams to reliably harmonize, label, and export multimodal time-series data with full provenance—bridging the gap between raw data, analytics, and deployable machine-learning systems in real-world environments.
Press Inquiries
Jerry Louis-Jeune
jerry@sophelio.io
https://sophelio.io
2008 Holland Ave #B,
Austin TX 78704
A video accompanying this announcement is available here: https://youtube.com/watch?v=BYPinRn4d8I
