Research & Publications

Advancing the State of the Art
in Enterprise AI Systems

Peer-reviewed research, scholarly articles, and technical publications contributing original findings to the fields of enterprise AI, multi-agent systems, and production machine learning at scale.

Selected Publications

Peer-Reviewed Research & Articles

Published work spanning academic journals, conference proceedings, and industry research venues.

Towards Reliable Multi-Agent Orchestration in Production Enterprise Systems: Patterns and Anti-Patterns
Javalkar, V. et al. ·
This paper examines production deployment patterns for multi-agent systems in enterprise environments, cataloguing recurring coordination patterns and critical failure modes observed across enterprise deployments. Introduces formal models for agent state management and proposes reliability primitives for autonomous agent networks achieving high task completion rates in production.
Featured Multi-Agent Systems Enterprise AI Distributed Systems
Scaling Retrieval-Augmented Generation: Architectural Trade-offs for High-Throughput Enterprise Deployments
Javalkar, V. et al. ·
A systematic study of architectural choices in RAG systems at enterprise scale — analyzing latency-quality trade-offs across retrieval strategies, re-ranking approaches, and context management techniques. Presents a cost model and decision framework for selecting RAG architectures based on query volume, latency SLAs, and accuracy requirements.
RAG Information Retrieval LLM Systems Performance
ML Platform Maturity Dimensions: A Framework for Assessing Enterprise Machine Learning Readiness
Javalkar, V. et al. ·
Presents a comprehensive maturity framework for enterprise ML platforms, derived from assessments of organizations across financial services, healthcare, retail, and technology sectors. Defines maturity dimensions, capability indicators, and evidence-based advancement pathways validated through longitudinal study across diverse enterprise contexts.
ML Platform Enterprise Empirical Study
Cost-Optimal Inference Routing for Large Language Models in Multi-Tenant Enterprise Environments
Javalkar, V. et al. ·
Presents an adaptive routing algorithm for LLM inference that dynamically selects models based on task complexity, latency requirements, and cost constraints in multi-tenant settings. Demonstrates significant cost reduction while maintaining accuracy parity with fixed model routing strategies, evaluated across production enterprise workloads.
LLM Optimization Inference Cost Engineering
Governing AI in the Enterprise: A Risk-Stratified Framework for Responsible LLM Deployment
Javalkar, V. et al. ·
Proposes a risk-stratified governance framework for enterprise LLM deployment, classifying use cases by risk profile and prescribing commensurate oversight mechanisms. Includes audit trail architectures, human-in-the-loop patterns, and incident response playbooks derived from real-world enterprise AI governance programs.
AI Governance Responsible AI Risk Management
Feature Store Architecture Patterns for Real-Time Machine Learning in High-Frequency Trading Environments
Javalkar, V. et al. ·
A detailed examination of feature store architectures optimized for ultra-low-latency ML inference in financial applications, covering point-in-time correctness, sub-millisecond feature retrieval, and consistency guarantees required for regulatory compliance in high-frequency trading contexts.
Feature Store FinTech Real-Time ML
Research Interests

Active Research Areas

Ongoing and emerging research directions at the intersection of enterprise systems and AI.

◈
Agentic System Reliability
Formal methods and empirical approaches for ensuring reliability, safety, and predictability in autonomous multi-agent systems deployed in production enterprise environments.
⬡
LLM Inference Optimization
Hardware-software co-design approaches for reducing cost and latency of large language model inference at enterprise scale without sacrificing output quality.
◎
AI Governance Frameworks
Technical mechanisms for operationalizing responsible AI principles — bridging the gap between policy intent and production system behavior at enterprise scale.
◉
Enterprise Knowledge Graphs
Integrating structured enterprise knowledge with neural language models to improve accuracy and traceability in high-stakes domains requiring auditability.