embedded world Product Showcase: Aurora Labs’ AI-Driven LOCI (Line-of-Code-Intelligence) Platform
March 03, 2025
Sponsored Story
The world of software development is advancing rapidly, and today’s developers are responsible for ensuring the reliability, quality, flexibility, and overall efficiency of modern, advanced systems. But managing and monitoring a wide variety of system insights across the software lifecycle can be overwhelming without the proper assistance.
The LOCI (Line-of-Code-Intelligence) Platform from Aurora Labs is designed to assist developers at each step of the development processes, providing intelligent insights and observations on static binary files.
The LOCI Platform in Action
The AI-driven platform is powered by Aurora Labs’ proprietary vertical LLM, known as Large Code Language Model (LCLM), specifically designed for analyzing compiled binaries. The LCLM is capable of performing binary analysis and providing insights into software behavior changes with its minuscule vocabulary and six GPUs. The LCLM also enables dynamic software analysis.
LOCI also models compiled binaries with real-world data to assist with testing and validation, test coverage references, detecting anomalies, predicting performance impacts, detecting deterioration across software versions, and identifying power-consuming functions. For early optimization, LOCI also integrated directly into CI/CD pipelines for continuous monitoring, early anomaly detection, and reduced rollbacks/hotfixes.
For AI inferencing and managing AI infrastructure, the LOCI platform features a Reliability, Availability, and Serviceability (RAS) solution for in-field analytics leveraging Deep Neural Network (DNN). This is designed to assist with in-field device analytics for AI systems with benefits like:
- Silent Data Corruption (SDC) Detection: Monitors PVT, ECC, GPU, CPU, and memory integrity, offering root cause analysis for corruption issues.
- In-Field Real-Time Monitoring: Predicts degradation in power, temperature, performance, and quality, enabling proactive management.
- Temperature Behavior Analysis: Detects anomalies in specific dies and cores, pinpointing affected code for precise resolution.
- Voltage Optimization: Provides voltage adjustment recommendations based on ECC increases and system performance.
- System Insights and Performance Monitoring: Predicts workload trends, detects bottlenecks, optimizes cold startups, tracks event deviations, and root cause analysis for improved reliability.
- Specific Data Insights: Identifies issues like missing database entries, module inconsistencies, and code-specific problems down to the line and core level.
Getting Started with LOCI
Today, the LOCI platform supports hardware from suppliers such as Infineon, NVIDIA, NXP, Qualcomm, Renesas, Samsung, and STMicroelectronics, as well as improving software reliability and predictive maintenance for cloud, mobile applications, high-performance computing (HPC), and embedded systems.
LOCI 2.0 extends the capabilities of tools like GitHub Copilot by advising on quality, reliability, and compatibility issues post-coding, ensuring alignment with defined KPIs.
LOCI does not require any agents or access to internal systems. Users can upload a binary file to receive specific contextual insights.
LOCI is currently only C, C++, and Go software running on ARM / AURIX processors is supported.
Additional Resources:
