Bhaskar Anjappa

Gen AI Engineer

ARYA

Jul 2023 - Present • 3 yrs 1 mo

Generative AI Engineer @ Aarya Tech Solutions (July 2023–Present) Built and deployed production-grade RAG (Retrieval-Augmented Generation) pipelines using LangChain and Milvus, improving semantic search retrieval accuracy by 35% on proprietary datasets. Designed hybrid search systems combining dense vector embeddings with keyword search to solve retrieval failures common in long-document technical Q&A. Architected and shipped 10+ AI automation workflows using n8n and Make.com, integrating disparate business tools (CRMs, email, calendars, task managers) to eliminate manual operational overhead for internal teams. Developed and deployed 10+ Custom GPTs and Google Gems for internal process optimization, using advanced prompting strategies (ReAct, Chain-of-Thought, Few-Shot) to reduce hallucination rates and improve tool-calling reliability in production agent workflows. Built high-performance RESTful APIs using FastAPI to serve LLM inferences at sub-200ms latency for client-facing applications — engineered for production traffic, not demo-scale. Integrated LangSmith for full observability and tracing across multi-step agent workflows, enabling systematic debugging of complex reasoning chains and token-usage optimization. Leveraged Google Cloud's Vertex AI and Gemini models for domain-specific prompt tuning, automating text summarization and classification tasks previously done manually. Independent projects: Built a hierarchical multi-agent research system (Planner-Researcher-Writer pattern) using LangGraph's cyclical state management — agents autonomously plan, execute, and critique in feedback loops. Built a GPT-style transformer from scratch in PyTorch (25M parameters), implementing multi-head attention and layer normalization to deeply understand transformer internals beyond framework abstractions.