Shaping automotive embedded computing: From real-time control to edge AI
Expert Perspective: Manish Taparia, Product Manager DFA Embedded Compute Platform
What is at the heart of your role as Product Manager for the DFA Embedded Compute Platform?
As a Product Manager in Bosch's semiconductor business, I am responsible for shaping the product strategy as well as the technology roadmap and the commercialization of our Data Flow Architecture (DFA) embedded compute platform. DFA combines deterministic floating-point hardware acceleration with embedded software, development tools, and integration support. I translate customer and market needs into product requirements that guide semiconductor architecture and investment decisions. My goal is to develop competitive solutions that meet automotive computing needs and create long-term business opportunities.
Automotive computing is changing rapidly. What developments are having the greatest impact on your work?
The shift toward software-defined and electrified vehicles is increasing the need for onboard computing power, from real-time control to edge AI. This is driving heterogeneous architectures in which general-purpose processors work with specialized computing resources. At the same time, hardware performance alone is no longer enough. Software, development environments, and system integration are essential, so product management increasingly focuses on integrated computing solutions rather than individual semiconductor capabilities.
When shaping the DFA platform, what are the key decisions you have to make?
I evaluate customer requirements, prioritize computing capabilities, and align technical and commercial decisions with our long-term product strategy. A central question is how DFA should scale across automotive domains and which workloads benefit most from specialized acceleration. Together with engineering and business stakeholders, I balance architecture and software readiness, development effort, customer value, and commercial potential to define product requirements, investments, and roadmap priorities.
What fascinates you most about turning a computing architecture like DFA into an automotive solution?
What fascinates me most is connecting semiconductor technology with automotive applications and business strategy. Developing an embedded compute platform means bringing architecture, software, system integration, and customer requirements together. I enjoy working across disciplines to turn complex technical challenges into clear product opportunities. With DFA, it is especially rewarding to see an innovative architecture evolve into a broader solution for different automotive applications.
Where can DFA make the biggest difference for automotive customers?
Customers need more computing performance, predictable execution, and efficient use of system resources. DFA complements CPUs in heterogeneous MCU and SoC architectures by accelerating suitable floating-point workloads. Offloading time-critical mathematical operations to dedicated hardware helps customers use processing resources more effectively while maintaining predictable execution. Supporting software and development tools simplify evaluation and integration, while my role ensures that the platform remains aligned with real application requirements.
Can you give us a concrete example of how DFA can help solve a real automotive computing challenge?
Successful projects require the right computing capability and efficient system integration. Motor control is a good example: Electric powertrain algorithms must perform complex calculations within strict real-time constraints. DFA can accelerate suitable computations with predictable timing and reduce the CPU load. To support adoption, we combine semiconductor IP with embedded drivers, code-generation tools, integration guidance and development resources. By incorporating customer feedback into product planning, we can prioritize practical engineering challenges and reduce implementation complexity.
Looking ahead, what will define the next generation of automotive embedded computing?
Three trends are shaping automotive embedded computing: heterogeneous computing, Edge AI, and architectural scalability across automotive domains. Intelligent functions are moving closer to sensors, actuators, and controllers. Hence, platforms must handle diverse workloads with demanding requirements for performance and latency as well as predictability and energy efficiency. CPUs and specialized accelerators therefore need to work together, with each workload running on the resource best suited to it. For DFA, we aim to scale a common architecture from deterministic real-time control and signal processing in powertrain and chassis toward more demanding workloads in centralized powertrain controllers, advanced chassis control, ADAS, radar, and sensor fusion. Future differentiation will not simply depend on adding processing power, but rather on combining specialized architectures, intelligent software, and application-driven innovation.
Bringing a platform like DFA to market requires many perspectives. How do you bring them together?
Competitive semiconductor solutions require close collaboration across system architecture, hardware and software engineering, functional safety, product management as well as customer-facing teams. For DFA, I connect application requirements with technical capabilities, development priorities, and product strategy. We also work with international technology partners and customers whose feedback helps us understand integration challenges and future requirements. Aligning technical expertise, customer needs, and business objectives is essential to turning complex technologies into competitive solutions.


