SOS platform helps to address the semiconductor industry’s data ‘emergency’
An interview with Pedro Pires, AI Product Manager at
Keysight Technologies and
PA: What is SOS Enterprise, and how does it extend the capabilities of Keysight’s existing SOS Core platform?
PP: SOS Enterprise is Keysight’s design data management platform built for global, multi-site engineering organizations that require secure collaboration, governance, and AI-ready infrastructure at scale. It extends SOS Core – which delivers high-performance version control, seamless EDA integration, and metadata-driven IP reuse for small to mid-sized teams – by unifying distributed engineering environments under a single, traceable system of record.
Where SOS Core focuses on streamlined, reliable data management for departmental or single-site teams, SOS Enterprise adds enterprise-grade collaboration, and end to end IP lifecycle traceability. It connects data across sites and domains, catalogues and governs IP for reuse, manages bills-of-materials and workspaces, and prepares engineering data for analytics, machine learning, and agentic AI workflows.
In short, SOS Core empowers teams to work efficiently and avoid data conflicts; SOS Enterprise transforms that operational data into a strategic asset that powers innovation at organizational scale.
PA: Why is engineering data governance emerging as a critical bottleneck for AI adoption in semiconductor design?
PP: AI / ML pipelines depend on structured, accessible, and trustworthy data. Yet the reality across most semiconductor organizations is that design data remains largely unstructured and siloed, making it unusable without substantial engineering effort, that often does not scale well.
Without standardized metadata, lineage, and well-defined APIs, teams cannot leverage historical design data for predictive analytics, yield optimization, or design automation. Manual, inconsistent governance further exposes data to security risks and compliance gaps. There is no clear path from raw engineering archives to curated, labeled datasets that ML models require.
Moreover, semiconductor companies are often subject to very stringent regulations, which require assets to be traceable to the root inputs and experiments to be fully reproducible, years after the set up.
This is why governance is often seen as the bottleneck: until organizations can catalog, version, and contextualize engineering data – and ensure that data is discoverable and traceable – AI remains aspirational rather than operational. Platforms that solve the governance layer solve the AI-readiness layer simultaneously.
PA: What specific challenges do fragmented design files, verification data, and IP libraries create for engineering teams today?
PP: The challenges are multi-dimensional. First, fragmented toolchains and isolated data prevent automation and insight across design, verification, and test. Without a unified data model, cross-domain traceability is impossible, and engineers cannot reliably determine which design version generated a given simulation result.
Second, inconsistent versioning, missing lineage, and weak dependency tracking cause rework and re-spins. Design knowledge gets trapped in silos, metadata is inconsistent, and IP blocks become hard to find. Engineers repeatedly recreate existing blocks, and organizations lose institutional knowledge.
Third, the productivity cost is substantial: Legacy storage systems slow down data operations, version conflicts arise when multiple engineers work on the same block, and teams must rely on manual coordination across global locations. For engineering leaders, this translates into limited visibility into project progress, data utilization, and compliance adherence – forcing reactive rather than proactive decision-making.
Taken together, these challenges create a compounding drag on engineering velocity that grows worse as design complexity and team distribution increase.
PA: How does SOS Enterprise establish a “single system of record,” and why is this important for enterprise-scale semiconductor development?
PP: SOS Enterprise establishes a single system of record by unifying distributed engineering environments – spanning analog, digital, RF, and system domains – under one traceable data platform. It does this through real-time multi-site synchronization, and deep integration into EDA tools and workflows. Every artifact, dependency, and file is tracked for reproducibility and confidence, from schematic through to system-level integration, and all of them live in this single source of truth substrate that connects all steps of the development lifecycle.
For enterprise-scale development, this is essential for several reasons. Global design teams suffer from slow sync, version drift, and lack of visibility across projects and sites, all of which force manual coordination. There is also the challenge of knowledge accessibility and discoverability. It is not uncommon that redundant work is performed due to the lack of awareness that a given IP was already developed by someone else in the company. A single system of record eliminates these inefficiencies by providing a knowledge platform, consistent workspaces and global visibility. It also establishes the metadata-rich foundation required for governance, auditability, and compliance – requirements that are now board-level concerns in regulated markets such as automotive, aerospace, and defense.
Beyond operational efficiency, a unified system of record converts historical project archives into reusable, machine-readable datasets, laying the groundwork for AI-driven design workflows.
PA: Can you explain how automated data lineage and traceability directly improve AI model reliability in chip design workflows?
PP: AI model reliability hinges on knowing exactly what data a model was trained on, where that data came from, and how it has changed over time. In chip design, this means tracing the full chain from a specific design version, through the simulation parameters that produced a dataset, to the model that consumed it.
SOS Enterprise captures this complete design-to-simulation lineage automatically. Structured metadata and versioning ensure that datasets fed into ML pipelines are accurately labeled, contextualized, and reproducible. If a model produces unexpected results, teams can trace back to the exact design revision and simulation conditions that generated the training data.
This traceability also supports continuous learning loops. As design data evolves, model retraining can be automated with confidence that the underlying data reflects known, auditable changes rather than uncontrolled drift. The result is AI models that are not only more reliable, but also explainable and auditable – a critical requirement in industries where regulatory compliance governs design decisions.
PA: What role does SOS Enterprise play in helping organisations meet increasingly complex compliance and regulatory requirements?
PP: SOS Enterprise delivers compliance-ready design data management with built-in governance capabilities. This includes enterprise-grade role-based access control, encryption, end-to-end audit trails, and automated approval workflows.
For organizations in regulated industries – automotive (ISO 26262), aerospace and defense (ITAR, DO-254), healthcare (medical device regulations), and government programs – SOS Enterprise provides the lineage, auditability, and secure collaboration needed to satisfy compliance mandates without manual overhead. Design data access can be controlled by geography, project, and role, and every action on the data is logged and traceable.
This positions governance not as a bottleneck, but as a strategic accelerator: it is the reason global enterprises can standardize on SOS for controlled collaboration and audit confidence, reducing the risk of data loss, IP breaches, and certification delays.
PA: How does the platform address the needs of distributed engineering teams operating across multiple sites and geographies?
PP: SOS Enterprise is architected from the ground up for multi-site, global collaboration. It delivers real-time synchronization across sites, ensuring that every workspace reflects the latest design state. Its Smart-Cache technology creates distributed networks of data and user workspaces that remain lightweight regardless of project data size, minimizing data duplication while maximizing performance.
For distributed teams, this means no more version drift, no more slow syncs, and no more manual coordination across time zones. Engineers at any site work against the same, consistent system of record. Access controls can be configured per site, per role, and per geography to satisfy regional compliance and data sovereignty requirements.
The platform also integrates with corporate IT systems – SSO, LDAP, and PAM – ensuring that multi-site deployment aligns with existing enterprise infrastructure rather than creating a parallel administration burden.
PA: In what ways does SOS Enterprise reduce manual engineering workflows, and what measurable productivity gains can users expect?
PP: SOS Enterprise reduces manual workflows at multiple levels. Version control and check-in/check-out operations are automated and integrated directly into EDA design environments – so engineers can manage data without leaving their design tools. Metadata capture, lineage tracking, and dependency management happen automatically rather than through manual documentation.
Approval workflows, release tagging, and IP publishing follow standardized, automated processes that replace ad-hoc email and spreadsheet-based tracking. For CAD managers, pre-built configuration templates and centralized administration eliminate per-project manual setup.
In terms of measurable outcomes, customers can expect fewer re-spins caused by data misalignment, faster tapeout cycles through reliable IP reuse and shorter verification cycles through traceable design-to-simulation linkage. Collectively, these gains translate into shorter time-to-market and lower engineering cost per project.
PA: How does improved IP reuse across teams and locations translate into faster development cycles and reduced costs?
PP: Enterprises lose significant engineering effort annually recreating existing design IP. The root cause is that data is stored but not organized or discoverable – metadata is inconsistent, naming conventions vary by team, and there is no centralized catalog of validated blocks.
SOS Enterprise transforms past work into future value through a metadata-driven architecture that captures design lineage and enables searchable, governed IP repositories. Engineers can discover proven IP blocks, understand the validation status and dependencies, and consume them through standardized workflows. This turns the IP library from a passive archive into an active productivity multiplier.
The impact on development cycles is direct: reusing a validated analog block or RF front-end module avoids weeks of re-design and re-verification. At scale, across dozens of projects and hundreds of engineers, this compounds into measurably shorter tapeout timelines and reduced engineering costs. The platform also enables variant creation and configuration management, so teams can adapt existing IP to new process nodes or design constraints without starting from scratch.
PA: What security mechanisms are built into SOS Enterprise to protect sensitive semiconductor IP, particularly in regulated industries like aerospace, defence, and automotive?
PP: SOS Enterprise provides a comprehensive security framework designed for the most demanding environments. At its foundation is granular role-based access control – which governs who can access, modify, and distribute design data, configurable by project, site, geography, and organizational role.
In addition, data in transit is protected through encryption, while end-to-end audit trails record every access, modification, and distribution event to provide the forensic visibility needed for compliance reviews and security investigations. Automated approval workflows also enforce policy gates before IP can be released or shared, helping prevent unauthorized distribution.
For aerospace and defense customers, this supports compliance with frameworks such as ITAR and internal classified-data handling protocols. For automotive, it aligns with ISO 26262 traceability requirements. The platform’s on-premises and sovereign deployment options ensure data sovereignty for organizations that cannot place engineering data in shared cloud environments.
These mechanisms collectively protect IP not just from external threats, but from the internal risks of process non-compliance, accidental data exposure, and ungoverned collaboration.
PA: Have early customer deployments revealed any quantifiable benefits in terms of project orchestration, operational efficiency, or AI integration?
PP: SOS’s architecture is trusted by some of the world’s leading semiconductor companies for design data management at enterprise scale. Customers consistently report reduced design rework through reliable version control and dependency tracking, faster tape-out timelines enabled by IP reuse and cross-site collaboration, and improved audit readiness that streamlines compliance certification processes. Even at the level of discoverability alone, companies can save a significant amount of time.
On the AI readiness front, the industry is slowly converting data infrastructure to build up to fully agentic workflows and we are yet to know of specific performance KPIs measured on this topic.
PA: Looking ahead, how do you see the role of data management platforms evolving as AI becomes more deeply embedded in semiconductor design and verification?
PP: Data management platforms will evolve from operational infrastructure into the strategic intelligence layer of semiconductor engineering. The platform will become the foundation on which AI copilots, predictive analytics, and autonomous design agents operate.
This evolution is already underway. As AI models move from research to production in EDA workflows – optimizing placement, expanding verification coverage, predicting yield, and eventually driving agentic automation of entire design sub-flows – the dependency on structured, governed, and traceable data will only intensify. The organizations that invest now in organizing engineering data, building lineage, and exposing it through standardized APIs will be the ones able to capitalize on each successive wave of AI capability.
For Keysight, this is the core of our roadmap: connecting the data management platform to AI/ML pipelines, enabling the transition from design data to design intelligence, and ensuring that every dataset, model, and experiment is catalogued and reusable. The semiconductor industry is entering an era defined by data, collaboration, and intelligence, and the data platform is the foundation upon which all of that rests.
