Applied AI Scientist

The application window is expected to close on:

** Job posting may be removed earlier if the position is filled or if a sufficient number of applications are received** .

** Meet the Team*
* Splunk, a Cisco company, is building a safer, more resilient digital world with an end‑to‑end, full‑stack platform designed for hybrid, multi‑cloud environments. Join the Foundational Modeling team at Splunk, where we advance the state of AI for high‑volume, real‑time, multi‑modal machine‑generated data - including logs, time series, traces, and events. We combine deep AI research expertise with the scale and operational excellence of Splunk and Cisco's global engineering capabilities.

Our work spans networking, security, observability, and customer experience - designing and deploying foundation models that enhance reliability, strengthen security, prevent downtime, and deliver predictive insights across Splunk Observability, Security, and Platform at enterprise scale. You'll be part of a culture that values technical excellence, impact‑driven innovation, and cross‑functional collaboration - all within a flexible, growth‑oriented environment.

** Your Impact*
* + Contribute to the research, design, and development of large-scale foundation models for machine-generated data, with a primary focus on graph data and additional support for logs, time series, traces, and event modalities.

+ Develop and enhance distributed training and inference workflows, leveraging data-driven approaches to improve model quality, scalability, and operational efficiency.

+ Collaborate with engineering, product, and data science teams to understand requirements, incorporate stakeholder feedback, and deliver AI/ML solutions that address business and technical needs.

+ Share emerging ideas, technical insights, and best practices with teammates, contributing to technical discussions and helping advance team capabilities and project outcomes.

+ Explore and evaluate new AI/ML techniques, tools, and methodologies, applying relevant innovations to improve workflows, solve technical challenges, and support the team's roadmap and objectives.

+ Take ownership of assigned projects and deliver high-quality results with urgency, while proactively identifying obstacles, driving resolution of technical issues, and continuously improving development processes.

*
* Minimum Qualifications:

*
* + Master Degree in Computer Science, or related quantitative field, plus 2+ years of industry research experience.

+ Proven track record in at least one of the following areas:
Large-scale graph representation learning and Graph Neural Networks (GNNs) (e.g., GCN/GAT/GraphSAGE, heterogeneous GNNs, graph transformers), large language modeling for structured and unstructured data, multi-modal fusion of graph, text, log, and time-series data.

+ Solid proficiency in Python and deep learning frameworks (e.g., PyTorch, Tensor Flow)

+ Experience translating research ideas into production systems.

*
* Preferred Qualifications:

*
* + Deep experience with graph representation learning, graph transformers (e.g., GCN/GAT/GraphSAGE), spatio-temporal GNNs, heterogeneous graphs (HGNN/Relational GNNs), and knowledge-graph-augmented modeling.

+ Expertise in constructing and operating on large-scale graphs (entity graphs, service dependency graphs, topology graphs, causal graphs, or log-event graphs).

+ Hands-on experience with frameworks such as PyTorch Geometric (PyG), DGL, Graph Gym, GraphML systems, or custom GNN runtimes.

+ Advanced Anomaly Detection with Graph:
Track record developing hybrid graph-temporal approaches (e.g., GNN + Transformer, graph contrastive learning, dynamic graph forecasting) for detecting anomalies in high-volume operational data.

+ Hands-on experience developing, fine-tuning, or adapting foundation models for domain-specific data such as logs, time-series, graphs, operational telemetry, or enterprise knowledge, including representation learning across structured and unstructured modalities.

+ Large‑Scale Training & Optimization - Experience optimizing model architectures, distributed training pipelines, and inference efficiency to minimize cost and latency while…