Sumit Rakesh

About me

Sumit works as a Research Engineer with EISLAB Machine Learning and CDT at Luleå University of Technology (LTU), Sweden. His main research interests span: Hybrid Edge–HPC Testbed Design & Integration, Multi-tenant Resilience, Workload Placement across the Compute Continuum, and Multimodal ML & Reproducible Experimentation. 

Research focus

• Research focus (Platform): Designing continuum-ready multi-tenant testbeds and turning platform choices into measurable outcomes—repeatability, isolation, performance, and operational reliability across edge ↔ on-prem/HPC.
• Research focus (Data/ML): Data-centric ML for noisy, multimodal signals (EEG, EEG–fMRI), including dataset creation and reproducible experimentation—useful for incident evidence/telemetry pipelines and benchmarking dependable methods.

Key areas across the Edge–HPC continuum & multimodal ML

• Continuum platform engineering: Architecting end-to-end Edge–HPC systems and heterogeneous GPU infrastructures, defining secure segmentation and routing boundaries, and designing fabric-aware connectivity under real-world integration constraints (power, physical layout, interoperability).
• Multi-tenant foundations: Designing and evaluating identity and access-control models (SSO/OIDC/LDAP-style), tenant isolation mechanisms, and least-privilege authorization patterns with auditable access guarantees.
• Workload placement: Designing and operationalizing placement strategies for different workload types (training and inference) across edge ↔ HPC, quantifying trade-offs in latency, data locality, and compute availability, and validating them through automated, policy-driven deployment (GitOps-style delivery + application traffic management for service exposure and routing).
• Data fabric: Designing and evaluating tiered storage for continuum workloads—high-throughput parallel POSIX storage (BeeGFS-style) for training I/O and checkpoints, complemented by S3-compatible object storage (MinIO/Ceph-style) for datasets and model artifacts; applying hot/warm/cold tiering based on access frequency and performance/cost trade-offs.
• Observability & Operations resilience: Building SRE-style telemetry and reliability foundations—metrics/logs/traces, SLI/SLO-oriented monitoring and dashboards (“golden signals”), and structured incident operations (runbooks, drills, post-incident learning) to improve MTTA/MTTR and reduce operational noise.
• Applied ML background: Optimization + ML for motion prediction, Signal processing & ML for EEG /multimodal neuroimaging.

Current projects

• Bothnia Edge–HPC Lab (EU co-financed) (2025–2028): Leading the end-to-end platform building and integration of LTU’s heterogeneous GPU Edge–HPC platform.
  Role: PI • Status: Ongoing

External Links

• Portfolio Github
• Publications of Sumit Rakesh (Google Scholar)
• Updated CV
• Open-Source Code Repositories at GitHub
• Updates at LinkedIn Profile
• Workshops