researcher = {
"name": "Ritesh Janga",
"role": "PhD Researcher in Computer Science (Data Science)",
"university": "University of Cincinnati",
"research": [
"Reliable and efficient LLM systems",
"Agentic AI and selective re-execution",
"Retrieval-augmented generation",
"Privacy-preserving and federated learning",
],
"mission": "Build AI systems that are useful, efficient, and worthy of trust."
}I work at the intersection of large language models, multi-agent systems, retrieval, and privacy-preserving machine learning. My research asks a practical question: how can intelligent systems remain accurate and auditable while using less computation and protecting sensitive data?
| Area | What I investigate |
|---|---|
| Reliable Agentic AI | Evidence-sensitive agent re-invocation, dependency-aware execution, failure analysis, and quality–cost trade-offs |
| LLM Evaluation | Reproducible benchmarks, calibration, evidence grounding, traceability, and statistical evaluation |
| RAG & GraphRAG | Hybrid semantic/keyword/graph retrieval, source attribution, multi-agent orchestration, and auditable outputs |
| Efficient LLMs | PEFT/LoRA, quantization, selective computation, and resource-aware inference |
| Privacy-Preserving ML | Federated learning, differential privacy, non-IID data, and privacy–utility analysis |
- Selective Agent Re-invocation: dependency-aware evaluation over 13,500 instances across medical, scientific, and legal domains.
- Enterprise GenAI: evidence-grounded workflows over a 47,000-word proprietary knowledge base designed to support 7,000+ R&D scientists.
- Auditable GraphRAG: semantic, keyword, and graph retrieval with specialist agents, persistent workspaces, and 100% source traceability.
- Federated Computer Vision: privacy-aware crop-disease detection using ConvNeXt, EfficientViT, and ResNet-18 under heterogeneous non-IID settings.
- Climate & Sustainability AI: leading an interdisciplinary research team investigating efficient language-model applications.
➡️ View my complete research profile on Google Scholar
I am interested in research collaborations involving reliable LLMs, agentic systems, RAG, efficient inference, federated learning, and responsible AI deployment. If your work overlaps with these areas, feel free to connect through LinkedIn or explore my portfolio.
Building AI systems that know when to reason, when to retrieve, and when not to recompute.