Consultant AI ML Across Locations Artificial Intelligence & Data AI Insights & Engagement

Deloitte India (South Asia) · Consulting & Professional Services

Built an institutional memory layer serving 30+ internal users by transforming enterprise documents and

operational data into a persistent conversational knowledge system.

Rearchitected a hybrid RAG search system into a state-machine design, achieving ∼80% reduction in cost per

query and ∼50% reduction in input tokens through architectural optimizations across 5 iterations.

Fine-tuned Snowflake Arctic Embed v2 using hard negative mining and Matryoshka loss on domain-specific data

(EPC, solar contracts), achieving R@1 = 0.629 (+8.4% over baseline), matching Gemini-level retrieval performance.

Built real-time Voice AI agents using Pipecat and WebRTC with streaming STT, LLM, and TTS pipelines,

implementing turn-taking, silence detection, and latency optimization for conversational context management.

Developed a reusable LangGraph agent orchestration framework enabling Agent2Agent communication, stateful

execution, dynamic task routing, and tool integration for enterprise workflows.

Architected a multi-agent contract intelligence platform (Supervisor, SME, Database, Asset agents) generating

structured outputs with human-in-the-loop validation, improving document processing reliability.

Deployed Qwen-14B with LoRA adapters and 8-bit quantization on vLLM (GCP) for real-time analytics, enabling

cost-efficient inference at scale while maintaining output quality.

Integrated MCP, REFRAG, mem0, and an LLM Wiki system to enable tool orchestration, persistent cross-session.

B. Tech / M. Tech; B.E / M.E; B.Sc. / M.Sc.; B. Com / M.com; MBA/PGDM; BCA/MCA; PHD

Microsoft Certified: Power Platform Functional Consultant

Microsoft Certified: Power Platform Developer Associate (PL-400)

Microsoft Certified: Power Platform Solution Architect (Preferred)

Microsoft Copilot Studio Certification (Preferred)