About

I am a software engineer and researcher focused on the intersection of manufacturing systems rigor and AI orchestration. My foundational background is in Manufacturing Science & Engineering (B.Tech, COEP Technological University), which profoundly shaped my engineering philosophy: high-stakes systems require explicit, auditable correctness boundaries, not just statistically plausible outputs.

Research & Engineering Focus

Currently, my technical focus narrows into three interlocking areas:

• LLM Agent Reliability: Building evaluation infrastructure where AI system outputs are verifiable. This manifests in my work on formal gate conditions and verification lattices.
• Small Language Model Orchestration: Composing multiple smaller, specialized models (like DeepSeek or Qwen pairs) rather than relying exclusively on massive general-purpose models.
• Defensible AI Infrastructure: Architecting systems—like the Professional Intelligence OS and SkillHub—that operate on deterministic logic rather than black-box heuristics.

Production Systems

I prefer to validate research through production deployment. My most extensively documented body of work involves building manufacturing AI systems where errors have tangible consequences:

• EBOM → MBOM Transformation Engine: A production-grade pipeline and published research addressing a critical gap in industrial ERP environments: converting noisy engineering bills of materials into ERP-safe manufacturing bills without silently corrupting production data. It utilizes a novel 3-class conflict classification system and formal production-safety gates.
• Procurement & Inventory Optimization: A live AI system automating multi-criteria vendor scoring and safety stock calculations, actively managing procurement decisions for physical manufacturing operations.
• Conversational ERP (Bahava): Connecting complex, structured ERP data to accessible conversational interfaces like WhatsApp, bridging the gap between dense technical systems and non-technical staff operations.

Operating Principles

Every system I build expresses the same underlying engineering value: that AI and automation systems operating on high-stakes data must have explicit, auditable, and testable correctness boundaries. I believe in prototyping quickly, instrumenting the system, simplifying the interface, and preserving the reasoning.