Venkata Sukumar
Gurugubelli

• Set the strategy and own delivery for ABM's AI Factory — an enterprise-wide transformation spanning four verticals (Finance, HR, Core Tech Services, and Sales & Marketing): an 18-workflow program with 6 in flight and the rest on the roadmap.
• Lead a 16-engineer cross-functional team (frontend, backend, data, and AI engineering) and own the full delivery lifecycle — business discovery, solution architecture, infosec review, cloud provisioning, and production — for systems serving a 100,000+ workforce.
• Directed an AI-powered finance reporting automation initiative that roughly halved manual effort across a corporate reporting cycle, and a multi-agent proposal platform coordinating several specialized agents to cut roughly two days of effort per proposal.
• Delivered an HR leave-of-absence automation system handling hundreds of requests weekly — response time cut from days to under a minute, with multi-state compliance and bilingual (English/Spanish) support.
• Partner with executive stakeholders on build-vs-buy, AI governance, and roadmap prioritization across the program.

• Delivered an enterprise GenAI solution that strengthened sales pipelines through AI-driven RFP generation.
• Architected enterprise-scale RAG pipelines — 65% faster query response; cut document processing time 85% via vector-index optimization.
• Processed 100K+ invoice documents monthly with multi-modal document-understanding models, surfacing substantial recoverable revenue.
• Fine-tuned quantized LLMs (Llama, GPT, LLaVA) and built batch extraction + real-time document Q&A products on Azure ML.

• Led NIH-funded research projects and mentored junior members in research methodology.
• Developed the iPredict neuro-fuzzy framework for outcome prediction on large-scale longitudinal datasets, improving precision and generalizability.
• Harmonized complex longitudinal dietary data from six study centers (200K+ rows, 200+ variables) using Python, R, SAS, and SPSS.
• Leveraged Azure/AWS to optimize processing pipelines and model training.

• Developed eFCM, a novel fuzzy clustering model for smoking-intervention outcome prediction, outperforming traditional clustering on accuracy and performance.
• Co-authored and presented the work at an international conference.

• Optimized interfaces through feedback analysis and quantitative analytics in Python; ran user interviews and usability tests.
• Built personas, journey maps, and prototypes in Sketch/InVision; reported UX KPIs to drive design iterations.

Dissertation: "iPredict: Intelligent hybrid neuro-fuzzy system for predicting outcomes in longitudinal trials."

Committee: Dr. Hua (Julia) Fang (chair), Dr. Ming Shao, Dr. Honggang Wang, Dr. Bruce Barton

Focus: data science, algorithms, data visualization (D3), and machine learning.