srinidhi shanmugavadivel

I’m an AI Engineer with hands-on experience building production-grade LLM systems, agentic pipelines, and RAG architectures. I deliver end-to-end projects that go beyond demos—into working, deployable products.

I’ve built multi-turn, document-aware RAG assistants with source citations and topic-switching to reduce hallucinations. I also created agentic tool-calling systems that support natural language queries across multiple uploaded sources and data connectors.

My work includes a document intelligence pipeline using a multi-engine OCR stack (Tesseract, PaddleOCR, EasyOCR) for extracting structured fields with confidence scoring and validation reporting. I’ve also extended an “Universal Data Connector” into modular FastAPI services with Dockerized, pluggable connectors and business-rule enforcement for freshness and rate limits.

I’m particularly excited by practical reliability and privacy. For example, I’m extending a privacy-preserving voice data assistant with voice-obfuscation (pitch-shift + time-stretch) while keeping transcription accuracy, and I’ve Dockerized on-prem deployments using Whisper with spoken, agentic responses.