Continuous Optimization of Agentic Workloads on Heterogeneous AI Systems
imec
Design the next-generation distributed runtime to orchestrate autonomous AI agents on heterogeneous compute platforms.
Optimized AI: algorithms, compute system and semiconductor technology
The Optimized AI team at imec explores how to design innovative AI compute systems and how best to implement AI applications on them, considering compute-system architecture, algorithm selection, and semiconductor technologies.
Dynamic workloads on multi-vendor multi-compute edge AI systems
Real-world AI systems, such as autonomous robots, are becoming increasingly capable by combining powerful compute platforms, diverse sensors, and components from multiple vendors. AI hardware is very diverse, adding different types of accelerators, going beyond traditional GPU and CPU architectures. At the same time, the software of these systems increasingly depends on agentic AI architectures, including multi-agent pipelines, LLM-based decision loops, and reactive tool-calling agents, to support real-world decision-making. Furthermore, the requirements of these workloads can change rapidly over time, driven both by the task being executed and by the information captured through the sensors. All of this requires an orchestration runtime that a) can continuously adapt the way workload is scheduled on the system depending on the agents' decisions, the environment and the AI task to executed, and b) a runtime that can take into account the best match between the currently running AI tasks and the available heterogeneous hardware components. [C(1]
Key Research Challenges & Objectives
This research will investigate tools and methods for the scheduling and placement of AI workloads on distributed edge compute systems. The developed algorithms should be -
Real-time aware : to take into account end-to-end latency and required time to action for the task -
Heterogeneous Hardware aware: Formulate placement algorithms that schedule agents and their tool invocations based on hardware topology, CPU/GPU loads, network connectivity, memory pressure, and model weights' physical locality -
Multi-objective aware: to find the best solution depending on real-time, energy, performance and memory requirements -
Workload aware: to make sure the scheduling and placement is adapted when the workload changes due to a change in AI task or a change in input data -
Future proof : to make sure the method and the implementation can be adapted to future workloads and compute hardware
[C(1] Not for the proposal, but Sourav dropped a student of his at my desk today who would like to pursue a PhD. He studied computer science, so Sourav thinks he may be a better fit for Optimized AI. I chatted a bit with him and he sounds good.
Required background : Computer Science, Electrical Engineering, or similar
Type of work : 60% software development, 20% experiments and simulation, 20% literature
Supervisor : Roel Wuyts (KU Leuven)
Daily advisor : Tom Vander Aa The reference code for this position is 2027-039 . Mention this reference code in your application
- ...memory requirements for agentic AI systems are growing rapidly.... ...deployments increasingly feature heterogeneous model ensembles—large... ...sophisticated software optimizations—paged attention,... ...Characterization of Multi-Agentic Workloads: This phase will focus on profiling and modelling...Suggéré
- ...Device heterogeneity in future semiconducting systems challenges design-technology co-optimization research to look beyond pinpointing a complex device architecture with an optimal... ...differences in technology within on application, addressing these trade-offs for...SuggéréStage
- ...Electronic systems are increasingly built from more than one technology: different process nodes, different... ...left to experience. The student will work on a partitioning and exploration framework for heterogeneous systems: cost models that describe each...SuggéréStage
- ...diversity of domains using AI/ML-based applications... ...growing diversity of workload requirements from electronic systems. While heterogeneous computing systems (CPU... ...provide some level of optimization, the static nature of... ..., CGRAs, etc.) focus on runtime...SuggéréStage
- ...Artificial Intelligence (AI) systems rely on advanced semiconductor technologies that are increasingly... ...failures. Furthermore, by leveraging the workload dependence of physical aging mechanisms, the project seeks to optimize the deployment of such monitoring and recovery...Suggéré
- ...artificial intelligence, machine learning, and data-driven process optimization. A master student pursing any of the above mentioned relevant... ...internship at imec, the candidate will develop and evaluate AI/ML models that can be integrated into our ETM process...Stage
- ...semiconductor technologies continue to scale, reliability... ...degradation, and workload-dependent operating conditions... ..., architectures, and system-level functionality.... ..., given the increasing AI chip power densities, the... ...embedded monitors and on-chip sensors. Utilize...Travail posté
- ...and high-capacity connectivity systems are increasingly moving... ...manufacturing processes can be optimized for operation beyond 100 GHz.... ...technologies for antenna-in-package and heterogeneous RF integration at sub-THz... ..., with an initial focus on the D-band (110–170 GHz). The...
- ...Understanding this behavior m odern AI training and inference workloads is crucial for designing optimized memory hierarchies and minimizing data‑movement bottlenecks... ...locality. By integrating these insights into a system‑level simulation environment, the student will evaluate...Stage
- ...times per day in pharmaceutical and petrochemical quality control on exactly such instruments. This project moves titration from... ...dosing reagent and waiting for an endpoint, the signal is read continuously: when the sample composition changes, the equivalence point moves...Temps partiel
- ...toepassingsgebieden zoals computertechnologieën & systemen, gezondheid, automotive, energie,... ...die positie als toonaangevende speler in ons vakgebied verder te versterken, zijn we voortdurend... ...In dit Training Program leer je alles on the job. Geen uren theorie, maar meteen...Temps completTravail postéTravail le weekend
- ...Bioinformatics & AI Expertise Unit VIB–KU Leuven Center for Cancer Biology (CCB), Leuven... ...) is a leading research institute focused on understanding cancer biology and... ...that support research at CCB. This includes agentic workflows that help researchers find suitable...Emploi contractuelTemps complet
- ...What you will do Continued technology... ...the integration of heterogeneous technologies are... ...management, and energy optimization across the entire... ...and optimize system behaviour beyond... ...architectural decisions, workload characteristics,... ...-level impacts on power, thermal,...
- ...design the future of AI computing What if... ...computing systems? At imec, you will... ...infrastructure is built, optimized, and scaled. This... ...of large-scale AI workloads. Developing... ...Belgium. You'll work on technology challenges... ...expertise and facilities, continuous learning through...
- ...novel 3D packaging and integration mechanism to stack logic, memory and I/O), physics simulations are required at the system-level to push optimization further across the stack. But as individual components shrink in size, this results in an impractical scale range for...
- ...our business to thrive and focus on empowering employee use of... ...be key in developing impactful AI solutions that drive business transformation... ...Architect and design scalable systems that can handle large volumes of data and complex AI/ML workloads, ensuring high performance and...
- ...Develop a system-level microarchitecture that integrates qubit arrays, cryogenic control electronics and classical co-processors. The work... ...their co-integration with the qubit chip. The study will focus on latency and signal integrity aware control, circuit thermal management...Stage
- ...therapeutics. Yet their analysis often depends on workflows in which separation,... ...well-defined protein mixtures towards more heterogeneous samples relevant to protein characterization... ...scientific question is how a compact fluidic system can distinguish, concentrate and route...Stage
- ...Organ-on-chip systems are emerging as powerful models to study complex biological functions... ...microenvironments. Recent advances have enabled continuous monitoring of tissue activity through... ..., cellular state, and tissue heterogeneity remain largely hidden. This PhD aims...
- ...System scaling constraints on power delivery call for enhanced power management through novel power component... ...switches and LDO. Exploring these heterogeneous power delivery applications requires cell-level models and optimization. In this PhD, you will assess various...
- ...Optimization of TEM Chunk Preparation Transmission Electron Microscopy (TEM) is one of the most powerful characterization techniques used... ...semiconductor research and manufacturing, providing detailed information on material structures and device features at the atomic scale. A...Stage
- ...looking for a Generative AI Engineer (GenAI) to... ...challenges with a focus on GenAI applications. You... ...RAG-based assistants, agentic workflows and automation... ...iterate your solution to continuously improve value... ...implementing, deploying and optimizing cloud-based AI solutions...Temps completEmploi en CDI
- ...solutions. When the project you work on also requires hardware... ...the selection and integration of optimal hardware components (cameras & lenses... ...in a high-impact industry. Continuous Learning – Access to advanced training, AI workshops, and deep tech development...Temps complet
- ...Visible Hybrid Photonic Systems The integration of... ..., this PhD will focus on the development of... ...integrated closed loop optimization of the optical power,... ...existing imec expertise in heterogeneous laser integration and... ...circuit design, requiring continuous exchange of models,...Stage
- ...we deliver end-to-end data and AI solutions, from strategy and platform... ...engineering and AI-enabled systems. That broad scope is reflected... ...ML, modern GenAI, and Agentic AI use cases. Your work goes beyond... ...AI solutions land, with a focus on product thinking and real business...À distance
- ...Love AI? Fascinated by security? Want to work on technology that will power the... ...machine learning, agentic AI security, confidential... ...AI. Secure AI workloads running on... ...Science, Embedded Systems, Cryptography, Electrical... ...actively support continuous technical and...Temps completTravail à domicile
- ...increasingly driven by heterogeneous integration, where... ...within a single system to deliver... ...approaches—such as wafer-on-wafer, die-on-... ...performance computing, AI, and mobile... ...insights into process optimization and quality control... ...supporting continued innovation in semiconductor...
- ...What you will do System architecture and trade-off analysis. Translate system-level requirements... ...Contribute to ongoing research projects on PIC-based optical interconnects, optical... ...environments. Familiarity with III-V or heterogeneous integration. Experience contributing to...Temps complet
- ...our entire product portfolio. We are looking for a System Security Architecture Expert who thrives on solving complex security challenges, thinks end‑to‑... ...Stay ahead of disruptive industry trends , including AI/ML, and explore the risks and opportunities they...Temps completTravail à domicile
- ...redistribution layers (RDL) on photonic interposers—... ...robust solutions for system-level integration onto... ...and compatibility with heterogeneous integration schemes. Such... ...applications in AI/ML hardware, optical I/... ...), design, simulation, optimization (30%), layout (10%), cleanroom...
Voulez-vous recevoir plus d'offres d'emploi ?
S'abonner et recevoir des offres d'emploi similaires à Continuous Optimization of Agentic Workloads on Heterogeneous AI Systems. Soyez parmi les premiers à postuler !

