ICICI Prudential AMC · Team Lead (3 Engineers) · Production AI
I architect and deploy production AI systems at scale: real-time Voice AI, hybrid vector search (RAG), autonomous tool-using agents (MCP), and enterprise customer support automation in regulated BFSI environments.
Production multimodal voice AI assistant built entirely solo and integrated into ICICI Prudential's iInvest mobile application across Android and iOS. Enables investors to query live portfolio valuation, mutual fund schemes, risk comparisons, and investment FAQs with 2–4s end-to-end voice latency.
Production semantic search engine across official financial documents, SIDs, KIMs, FAQs, and video transcripts. Combines Milvus dense vector search with BM25 lexical retrieval and cross-encoder reranking, achieving 95%+ precision and sub-second retrieval at scale.
End-to-end AI-powered email triage, knowledge retrieval, and CRM synchronization. Resolves high-volume customer inquiries automatically with 85%+ accuracy grounded in internal Business Process Manuals (BPM).
Automated audio speech analytics platform auditing ~8,000 customer support calls every business day. Employs Whisper transcription, speaker diarization, regulatory risk flags, and automated QA scorecards.
Standardized Model Context Protocol (MCP) server connecting foundation LLMs to live core banking APIs, CAMS databases, and 11+ automated n8n workflows with deterministic security guardrails.
From architectural discovery and rapid prototyping to production hardening, security guardrails, and enterprise integration.
Low-latency streaming conversational voice agents for mobile apps, web, and contact centers via LiveKit WebRTC.
Hybrid retrieval uniting Milvus dense vector embeddings with BM25 lexical token matching and reranking.
Automated email/ticket triage, policy grounding, and downstream CRM resolution saving hundreds of staff hours.
Model Context Protocol (MCP) integrations allowing LLMs to reason, validate schemas, and execute system actions safely.
Whisper audio transcription, diarization, compliance scoring, and automated QA for thousands of daily calls.
Transitioning legacy RPA bots into resilient cognitive AI pipelines integrated with modern APIs and n8n workflows.
Lead a high-performance team of 3 engineers across intelligent automation & speech analytics (~8,000 calls/day), while personally building the flagship i-Assist Voice AI solo end-to-end.
Spearheaded 25+ process optimizations and the migration/orchestration of 180+ automation bots toward cognitive pipelines.
Engineered mission-critical UiPath automations, Python services, and API connectors achieving up to 99% workload reduction.
Direct inquiries welcome for production AI engineering, consulting, or technical advisory.
Email: dhruvshipsofficial@gmail.com