AI Infrastructure Pipeline
Reliable infrastructure for asynchronous, multi-stage AI workflows in a national-laboratory environment.
- Institution
- Oak Ridge National Laboratory
- Role
- AI Platform Engineering Intern
- Timeline
- May — Jul 2026
- Status
- Summer 2026
FastAPI services orchestrating multi-agent reasoning, retrieval, generative models, and molecular diffusion workflows.
Complex AI workflows required multiple reasoning, retrieval, and generation services to operate reliably as one system.
Build modular FastAPI services with asynchronous task execution around Google CoScientist, multi-agent tournament reasoning, RAG, and molecular diffusion models.
Dhruva engineered the end-to-end orchestration pipeline, service APIs, caching layers, retry mechanisms, fallback strategies, and Dockerized environments.
The resulting infrastructure standardized dependencies and improved robustness across multi-stage, multi-environment workflows.
AI platform work depends as much on failure handling and service boundaries as it does on model capability.