Catalin Caldararu

• Connected technical execution, product, innovation, and Agile delivery.
• Originated and created from zero a 35-item product prototype portfolio (feasibility demos), and filed 22 patent applications as first-named inventor, ranking as the most active inventor at Dell's Romania site and in the top 10% company-wide.
• Defined the portfolio's strategy, running agent loops on-device across NPUs and GPUs, with cloud fallbacks. Used ONNX Runtime and ONNX Runtime GenAI with execution providers such as AMD Ryzen AI, Qualcomm QNN, Intel OpenVINO, NVIDIA CUDA, and DirectML.
• Owned the backlog, business cases, PRDs, and draft solution architectures. Originated the inventions from directly observed customer needs rather than assigned specs, and grew the effort into an R&D roadmap delivered across 4 phases, each scoped from the results of the last.
• Won 9 Principal Engineers as voluntary co-inventors, holding no reporting authority over them or vice versa, and extended the portfolio across 8 domains: Privacy, Security, Resilience, Efficiency, Personalization, Accessibility, Automation, and Hybrid Edge SDLC.
• Facilitated weekly design-thinking sessions that pressure-tested draft architectures and prototypes against domain constraints and available on-device accelerators. Evaluated more than 60 inventions with the team, descoping many of them myself.
• Pitched each invention to engineering and legal leadership, converting approved concepts into filings with 4 patent attorneys. Ran the effort as a self-managing team with no manager, reporting to a VP-level sponsor.

(Stats per IDiyas Public Statistics)

• Owned and built the function-calling SLM and ASR services across two silicon platforms in a Hybrid AI product featured at CES and Dell Technologies World, shipping to every Dell AI PC customer. Created both from zero, as the only Dell services of their kind on that hardware, running inference on NPUs and GPUs.
• Built the accuracy, latency, and safety test suites (MMLU, WER) that benchmarked every candidate SLM and ASR model ahead of shipping.
• Took on work across the wider program over time, contributing to services that spanned all four silicon targets (AMD, Qualcomm, NVIDIA, Intel).
• Delivered, with the team, an integration layer that let internal teams and customers add SLM and ASR to their AI Apps without building inference stacks.
• Shipped that layer as part of a toolkit that cut time from PoC to production by ~80% for on-device AI applications, kept prompts and audio on the machine, and removed recurring cloud inference costs.
• The integration layer reached production use within months of launch, adopted by IT professionals building Agentic Edge AI applications and by organizations with privacy-first requirements.
• Resolved defects and cross-team dependencies alongside hardware vendors and 5 partner Scrum teams (Testing, MLOps, DevOps, Consumers).
• Facilitated my team's Daily Scrum, upskilling teammates in on-device AI and the product it shipped into, and reporting progress and impediments to internal and external leadership.

• Was awarded an off-cycle, performance equity grant at 18 months, ahead of the formal promotion cycle.
• Built a custom Jira-like work management system and the underlying task model, replacing a workflow fragmented across the legacy monoliths.
• Owned the configuration service feeding downstream systems the parameters for the platform's specialized task types, and the data aggregation service giving external partners a single outbound interface to their own work management systems.
• Led the adoption of HashiCorp Vault and Dynatrace across 240 service instances (60 services, 4 products, 4 environments), eliminating secret sprawl through a shared library and bringing telemetry to 100% coverage, sustained thereafter.
• Took over the team's CI/CD for a year, raising pipeline coverage from 75% to 95%. Migrated 120 components (20 services, 10 front-ends, 4 environments), right-sizing resources to free capacity for later development without adding to the team's footprint.
• Supported 10 partner Scrum teams consuming my team's services, resolving cross-team dependencies and integration defects as a point of contact for downstream consumers and E2E testing.
• Facilitated the Daily Scrum for a 5-person team for a year, onboarding 7 developers of varying experience levels and upskilling teammates on our modules and the platform they shipped into.