Sohrab Namazi Nia (Spencer Namazi Nia)
he/him
Computer Science PhD candidate (ABD) at NJIT BDAL
- NYC Metropolitan Area
- Google Scholar
- ORCID
- ResearchGate
- Github
Talks & Presentations
Invited talks, poster presentations, workshop talks, and research forums
ASMR: Agentic Schema Generation for Ship Maintenance Report Writing
Presented ASMR, a modular two-agent framework that discovers compact, reusable schemas from historical ship maintenance narratives. The talk walked through the Field Generation Agent, which extracts semantic concepts via adaptive multi-granularity clustering, and the Structural Optimizer Agent, which uses reinforcement learning to trim and merge candidates into a compact, non-redundant schema — nearly doubling schema informativeness.
Lower-Bound Distance Queries under Partial Information
Co-presented our VLDB 2026 work on answering lower-bound distance queries in a metric-space graph where only a subset of edge distances is known — without relying on any black-box distance oracle. Discussions at the poster covered the query-based model and how the framework trades off preprocessing overhead, query time, and bound tightness with provable guarantees under metric assumptions.
Utility-Aware Human–LLM Agent Orchestration for Data Science Pipelines
Presented our research on a novel framework for agentic collaboration between humans and LLMs across the data science pipeline. The talk covered how to orchestrate human and LLM agents to maximize utility for downstream ML tasks, with live demos of the pipeline architecture.
PhD Research Presentation — Personalized Recommendation and Top-k Set Retrieval using LLMs
Presented PhD research to fellow PhD students and faculty at the Ying Wu College of Computing at NJIT. The talk covered the full arc of my dissertation: LLM-powered personalized recommendations, probabilistic package selection, and agentic human–LLM collaboration in data science pipelines.
