Fine-Grained Mapping of AI Training & Inference Workloads Across the Full Memory Hierarchy
imec
Understanding this behavior m odern AI training and inference workloads is crucial for designing optimized memory hierarchies and minimizing data‑movement bottlenecks. In this internship, the student will develop a fine‑grained workload‑to‑memory mapping framework that characterizes how representative AI workloads interact with the entire memory stack. This includes profiling layer‑level and operator‑level footprints, traffic patterns, reuse distances, and temporal/spatial locality. By integrating these insights into a system‑level simulation environment, the student will evaluate bottlenecks, identify optimization opportunities, and generate actionable guidelines for future memory hierarchy design targeting large‑scale AI systems.
Skills to stand out:
Solid understanding of memory subsystem Familiarity with AI training and inference workload characteristics Strong programming skills in C++ or Python Experience with performance modelling techniques
Exposure to system-level simulation tools or benchmarking frameworks is a plus (Ramulator, DRAMSys, DRAMSim, Gem5, …)
Type of internship : Master internship, PhD internship
Duration : 6-9 months
Required educational background : Computer Science, Electrotechnics/Electrical Engineering, IT, Nanoscience & Nanotechnology
Supervising scientist(s) : For further information or for application, please contact Khakim Akhunov ( Voir email sur chipmakerjobs.com ) The reference code for this position is 2026-INT-045 . Mention this reference code in your application. Imec allowance will be provided for students studying at a non-Belgian university.
Applications should include the following information:
resume motivation current study
Incomplete applications will not be considered
- ...Current memory systems and hierarchies are struggling to keep pace with modern workload demands, especially... ...imposed by generative AI models. On-chip,... ...to string memories such as charge... ...to application mapping, by using system... ...with full- and sub-system...SuggéréStageTravail posté
- ...distributed runtime to orchestrate autonomous AI agents on heterogeneous compute platforms... ...semiconductor technologies. Dynamic workloads on multi-vendor multi-compute edge AI... ...topology, CPU/GPU loads, network connectivity, memory pressure, and model weights' physical...Suggéré
- ...Artificial Intelligence (AI) systems rely on advanced semiconductor technologies that... ...failures, along with register file and SRAM memory errors. While in-situ monitors, error-... ...failures. Furthermore, by leveraging the workload dependence of physical aging mechanisms,...Suggéré
- ...design the future of AI computing What if your... ...will drive innovation across the entire computing stack... ...of large-scale AI workloads. Developing performance... ..., accelerator, memory, and system designs. Collaborating... ...large language model training or inference. Strong understanding...Formation
- ...Artificial Intelligence (Edge AI), Internet-of-Things (IoT) devices... ...is driving an urgent need for memory technologies that combine... ...systems require non-volatile memories capable of reducing power... ...critical for unlocking their full potential in future memory...SuggéréStageRecrutement immédiat
- ...versterken, zijn we voortdurend op zoek naar gepassioneerde talenten die het verschil willen maken! Functieomschrijving In dit Training Program leer je alles on the job. Geen uren theorie, maar meteen meedraaien in een hightech omgeving, mét begeleiding van ervaren...FormationTemps completTravail postéTravail le weekend
- ...What you will do The MITO (Memory Innovation & Technology Optimization) group brings together... ...HBF, embedded DRAM, DRAM on logic, block memories, ferro based memories,...). As MITO Group... ...needs. You will drive collaboration across teams, groups and departments, fostering...
- ...by delivering high-impact work across your area of expertise. We... ...be key in developing impactful AI solutions that drive business... ...includes data preprocessing, model training, validation, and deployment.... ...of data and complex AI/ML workloads, ensuring high performance and...Formation
- ...increasing diversity of domains using AI/ML-based applications, there is a growing diversity of workload requirements from electronic... ..., communication, and memory elements limits the level of utilization... ...reconfigurable technologies, across computation, communication,...Stage
- ...aging, interconnect degradation, and workload-dependent operating conditions. Current... ...propagate aging and degradation effects across the design hierarchy, from devices and interconnects to... ...models. In addition, given the increasing AI chip power densities, the coupling...Travail posté
- ...What you will do Collaborate with cross-functional teams in applications, technology, circuits and design (across imec) towards R&D of novel memory/storage solutions in a system. Conduct feasibility studies through performance modeling and prototyping to provide insights...Temps complet
- ...The compute and memory requirements for agentic AI systems are growing rapidly... ...and cost. Current inference frameworks employ... ...DRAM, 3D-stacked memories, ternary content-... ...of Multi-Agentic Workloads: This phase will... ...KV cache operations across diverse agentic serving...
- ...Inspired by recent literature on AI-Driven Semiconductor Test Data... ...for secure, private model inference. Requirements Master’s... ...preferred. Experience in already trained LLM models locally is... ...spatial data (heatmaps/wafer maps). How to Apply Please...FormationStageRecrutement immédiat
- ...AI is everywhere. What matters now is where it actually makes a difference. We are... ...cases Strong communication skills, across both business and technical audiences... ...meal vouchers, bonus,...) ~ Continuous training to build and extend professional, technical...Formation
- ...transistors and use this knowledge to make them more reliable for future memory applications. The project will combine advanced electrical... ...“why?”, is excited by open scientific questions, wants to learn across disciplines, and is motivated to turn unexpected experimental...
- ...consumption are often determined by the behavior of the memory subsystem. Memory access patterns generated by accelerator workloads can significantly influence DRAM latency,... ...aims to study memory behavior under realistic AI workloads. Integrating a DRAM simulator with an...Stage
- ...Dataroots, we deliver end-to-end data and AI solutions, from strategy and platform... ...perspectives, and a common way of thinking across disciplines. Within that team, you design... ...actively encouraged and supported. A training budget for individual and team learning...FormationÀ distance
- ...Developing hardware for AI is both costly and technically demanding. The software stack... ...that does not yet exist for emerging AI workloads? Fortunately, most AI applications are... ...focuses on simulating accelerator memory behaviour during distributed LLM deployment...Stage
- ...Who we are As Full Custom Analog Layout Engineer, you join the Design Services Custom Layout team. This team specializes in full-custom... ..., and RF circuits. We support customers and internal programs across the complete ASIC development lifecycle, from design...
- ...The evolution of the next -generation computing critically needs significant improvement in On-chip memory, particularly in the AI-driven demand. In such a scenario, 2D semiconductors align naturally both at the device and the system levels. This is due to their superior...
- ...have a degree in computer science or related field and/or specific training. You have at least 3 years of experience in a business or... ...industry. Continuous Learning – Access to advanced training, AI workshops, and deep tech development . Great Work-Life...FormationTemps complet
- ...informed decisions. Our field data collection teams are spread across Europe and North America, ready to accept new challenges. Project... ..., this data will be used to enhance one the most popular online maps in the world. The data collectors will be given specific...Temps complet
- ...billions of complex, nanostructured arrays of transistors that perform memory and logic operations. In order to enable smaller, faster, more... ...execute time-resolved spectroscopic reflectance measurements across a large wavelength range; from the ultraviolet to the visible...
- ...At Datashift, we’re looking for a Senior AI Engineer who excels in crafting challenging technical solutions that deliver business... ...solutions and tooling (e.g., RAG, agentic workflows, vector databases, fine-tuning) Experience with cloud platforms (AWS, Azure, or GCP) to...Temps completEmploi en CDI
- ...Company description A full-time PhD position is available for... ...aims to develop and validate AI models that combine information... ...tasks Complete a doctoral training programme and conduct original... ...disease and HCC. Develop, train and evaluate machine-learning...FormationTemps completStageRecrutement immédiat
- ...Abstract Increasing the density of 3D memory in future technology nodes requires device concepts... ...and device-level characterization to map conductivity, contact behavior, electrostatic... ...rigorous data analysis, and collaboration across metrology, integration, device and...Stage
- ...Love AI? Fascinated by security? Want to work on technology that will power the next... ...ideas into real technologies that can scale across automotive, industrial, IoT, and edge... ...preserving AI, and trustworthy AI. Secure AI workloads running on advanced semiconductor...Temps completTravail à domicile
- ...2027 The VIB Center for AI & Computational Biology... ...position comes with full salary and core funding... ...students and postdocs from across the world. Assignment... ...Leuven will offer language training. You are required to... ...also only 14 minutes by train from Brussels Airport which...FormationTemps complet
- ...and integration mechanism to stack logic, memory and I/O), physics simulations are... ...system-level to push optimization further across the stack. But as individual components shrink... ...from the limited availability of reliable training data when it comes to large-scale problems...Formation
- ...We are looking for a Generative AI Engineer (GenAI) to strengthen our Artificial Intelligence team at Datashift. In this role you... ...voice agents Maintaining a scalable and efficient codebase for training, testing, and deploying GenAI applications in production...FormationTemps completEmploi en CDI
Voulez-vous recevoir plus d'offres d'emploi ?
S'abonner et recevoir des offres d'emploi similaires à Fine-Grained Mapping of AI Training & Inference Workloads Across the Full Memory Hierarchy. Soyez parmi les premiers à postuler !

