Kanhaiya

AI Engineer with 1+ year of experience building production-grade automation systems on AWS using LLMs and the LangChain ecosystem. Specializing in RAG pipelines (Bedrock, pgvector), multi-agent orchestration (LangGraph, CrewAI), and scalable FastAPI backends deployed on AWS. Delivered systems reducing manual processing time by 80% through automated document extraction and semantic search at enterprise scale.

AWS Certified Cloud Practitioner with strong foundations in prompt engineering, tool calling, and LLMOps evaluation (LangSmith, ragas).

AI Engineer - GITAN AI

(2025-06)

• Built an end-to-end pipeline where an AWS Fargate bot auto-joins Google Meet, records audio, and triggers AWS Lambda processing via REST-style service boundaries.
• Integrated AWS Transcribe with speaker diarization (99%+ accuracy) and produced structured JSON transcripts for downstream processing.
• Implemented a RAG pipeline using Bedrock Titan Embeddings + pgvector HNSW index for semantic search, supporting enterprise-scale retrieval.
• Designed a multi-tenant architecture with 3-layer data isolation to support secure, scalable deployments.
• Architected a 3-agent LLM system using Claude Sonnet 4.5 for end-to-end property recommendation via WhatsApp, including agent orchestration and memory using conversation history.
• Built semantic search over property listings using S3 Vector DB and embeddings, enabling fast AI search across listings.
• Deployed a scalable FastAPI backend on AWS EC2, exposing REST APIs for ingestion, retrieval, and recommendation workflows.
• Automated financial data extraction from Bank Statements using Amazon Textract OCR, reducing manual claim processing time by ∼80%.
• Built structured parsing and validation with automated form filling deployed on AWS EC2, standardizing outputs for downstream systems.

Master's - Mathematics and Computing - NIT Allahabad (MNNIT) (2023 - 2025)