Sohrab
Namazi Nia
Agentic data management, RL-powered optimization, personalized recommendation, and human–LLM collaboration in data science pipelines
PhD Candidate (ABD) at NJIT BDAL · Research funded by NSF & ONR · Agentic AI · Personalized Recommendations · AI for Healthcare
Research Interests
Where I spend my time thinking
Agentic AI & Multi-Agent LLM Systems
Designing multi-agent LLM frameworks, progressively evaluating agentic workflows to surface potential failures, and using reinforcement learning to guide agentic systems.
Human–AI Collaboration in Data Management
Orchestrating human and LLM agents across data science pipelines, with customizable trade-offs between result quality and cost.
Autonomous & Multimodal Data Systems
LLM-powered data management, personalized recommendation and package selection over multimodal data, and agentic data lake ingestion.
Applied Domains
Where my research meets the real world
Finance
At Scalata.ai (ranked #19 in the AI Hot 100), built a reasoning-driven multi-agent system for automated credit analysis using LangGraph — reducing analyst computation time by 60%. Designed an agentic data lake supporting 24+ document types and VectorDB-backed RAG pipelines for retrieval over heterogeneous financial documents.
Healthcare
Proposed G-CLAHE, a novel medical X-ray image enhancement method combining global and local histogram equalization — achieving an order-of-magnitude improvement in image quality and a 17% improvement in diagnostic accuracy for chest X-ray abnormality detection. Published in IJPRAI 2024.
E-Commerce
Researched LLM-powered personalized recommendation over multi-modal product data. Designed a multi-agent probabilistic framework for top-k package queries — minimizing expensive LLM oracle calls while guaranteeing optimality. Achieved a 45% speedup and 15–20× cost reduction vs. state-of-the-art. Also developed PromptXplorer for personalized prompt refinement using RAG and clustering-based composite prompt generation.
About Me
Hey, I'm Sohrab — my preferred name is Spencer. I'm a Computer Science PhD candidate (ABD) at NJIT with experience in building production multi-agent LLM systems at Scalata.ai (ranked #19 in the AI Hot 100).
My research, funded by NSF and ONR, focuses on agentic human–AI collaboration in data management and autonomous data systems. I graduated top of my class from IUST (GPA 4.0) and have received competitive research funding at NJIT. Outside work, I'm passionate about astronomy, space exploration, and football (soccer).
Request CV — Contact on LinkedIn