The purpose of this role is to support the design, implementation, and evaluation of a measurement-informed, network-aware static placement framework (Static NOVA) for geo-localized multi-cloud GPU training. The researcher will assist in collecting real-world cloud network measurements, developing analytical cost and feasibility models, implementing optimization logic, and supporting simulation-based evaluation for a systems research study.
Master’s degree or PhD (completed or pursuing) in Computer Science Engineering,Information Technology, Networking / Distributed Systems.
Minimum 2+ years of Experience
Strong academic or applied background in networking research
Python (NumPy, Pandas, Matplotlib)
Networking tools: iperf3, ping, traceroute
Optimization tools or libraries (ILP solvers, OR-Tools, PuLP)
Cloud platforms (at least one of AWS, GCP, Azure)
Prior experience with systems research, cloud computing, or network performance evaluation
Experience conducting measurement-based experiments (e.g., benchmarking, performance profiling)
Familiarity with distributed systems concepts (latency, bandwidth, synchronization, throughput)
Experience working on simulation or analytical modeling projects
Prior publication experience (conference or journal) is a plus
Strong understanding of:
Network latency vs bandwidth trade-offs
Cloud pricing models (compute and network egress pricing)
Distributed machine learning communication patterns
Ability to:
Translate real-world measurements into analytical models
Design cost and feasibility constraints
Implement static optimization logic
Strong analytical and problem-solving skills
Ability to document methodology clearly for academic use
The freelance researcher will be responsible for the following tasks:
- Network Measurement
- Measure inter-provider bandwidth and latency
- Organize results into structured matrices for analysis
- Pricing & Resource Data Collection
- Collect GPU pricing and network egress pricing from cloud providers
- Document provider-specific constraints and assumptions
- Analytical Modelling
- Assist in modelling distributed training communication traffic
- Develop feasibility constraints based on bandwidth and synchronization limits
- Support compute and network cost modelling
- Static Placement Framework Implementation
- Implement a static, network-aware GPU placement optimizer
- Integrate feasibility constraints into the optimization logic
- Simulation & Evaluation
- Run simulation scenarios across varying model sizes, GPU counts, and pricing conditions
- Compare Static NOVA against baseline placement strategies
- Generate plots and summaries of results
- Research Documentation
- Maintain clean, reproducible code and experiment documentation
- Assist in preparing figures, tables, and methodology descriptions for a research paper
Contact Detail : 95661 33822
Job Types: Part-time, Freelance
Pay: From ₹10,000.00 per month
Benefits:
- Flexible schedule
- Work from home
Work Location: Remote
