Ravi Ranjan

AI Engineer with experience building production-grade Generative AI systems, multi-tenant conversational AI platforms, and hybrid retrieval architectures using LLMs, vector databases, and knowledge graphs. Skilled in RAG & GraphRAG pipelines, open-source LLM deployment using vLLM, and scalable AI backend development with FastAPI.

Python Developer (AI/ML) at Impacto Digifin Technologies Pvt. Ltd. (2024-11 – Present)

• Architected the AI backend for a multi-tenant conversational AI platform supporting enterprise knowledge retrieval, FAQ automation, and customer interaction workflows.
• Designed end-to-end RAG and GraphRAG pipelines using Milvus and Neo4j with hybrid semantic and graph-based retrieval workflows for contextual document intelligence.
• Developed async FastAPI-based APIs and orchestration pipelines for document ingestion, retrieval, chatbot interaction, and response generation.
• Integrated open-source LLMs using vLLM for high-performance inference and evaluated models including Qwen 2.5 VL, Gemma, Phi, LLaMA, and GPT-OSS based on latency, hallucination behavior, and response quality.
• Implemented retrieval optimization using embedding similarity search, top-K retrieval, and BGE reranker models to improve contextual response relevance.
• Built voice-enabled conversational pipelines integrating STT/TTS models for real-time speech interaction.
• Implemented evaluation workflows using RAGAS for response quality validation and hallucination assessment across retrieval pipelines.
• Collaborated with frontend and DevOps teams for Dockerized deployments, AI system integration, and production infrastructure setup.

Bachelor of Engineering in Electronics and Communication – Acharya Institute of Technology (2019 – 2023)