AI Engineer · 1.5 YOE

HARIVARSHAN

Think of me as a perpetual playground resident in the world of systems. I'm slightly obsessed with designing scalable AI architectures and wiring up robust, end-to-end pipelines — always hunting for the next big lesson hidden in the code.

✦ P.S.This portfolio is actually a product of me messing around with "AI vibe coding." I wanted to see if an AI could design a clean, solid website without it screaming "AI DESIGNED THIS" at the top of its digital lungs. Welcome to the experiment!

PythonFastAPINext.jsPostgreSQLDockerLangGraphGenerative AIRAGAgentic AI
Photo of HARI VARSHAN

Selected Work.

Professional

WISE — Web Integrated Search Engine

Architected a highly scalable hybrid RAG system handling over 15,000 corporate filings. Engineered a querying engine routing between vector search and NL-to-SQL, returning context-aware answers with low latency.

RAGQwenLlamaGPT-5.4-mini
Professional

AI-Based Corporate Filings Parser

Designed an intelligent document processing pipeline extracting structured business data via dynamic Pydantic definitions. Implemented an "LLM-as-a-Judge" node to continuously evaluate extraction accuracy.

PydanticGemini 3 FlashGemini 2.5 Pro
Professional

Podcast Generation Engine

Developed an automated B2B TTS pipeline transforming executive briefings into multi-speaker podcasts. Features contextual abbreviation expansion and artifact-free audio cloning.

Gemini 2.5 FlashVibe VoiceDemucs
Personal Project

Autonomous RFP & Tender Response Swarm

Architected a full-stack, multi-agent application to automate enterprise B2B proposal generation. Orchestrated a cyclical LangGraph swarm with a Pydantic state model to parse tender requirements and draft compliant technical responses autonomously.

PythonFastAPINext.jsPostgreSQLLangGraph

Experience.

AI Engineer @ Euroland

2024 – Present

WISE — Web Integrated Search Engine

Architected a highly scalable hybrid RAG system handling over 15,000 corporate filings. Engineered a querying engine routing between vector search and NL-to-SQL, returning context-aware answers with low latency.

RAGQwenLlamaGPT-5.4-mini

AI-Based Corporate Filings Parser

Designed an intelligent document processing pipeline extracting structured business data via dynamic Pydantic definitions. Implemented an "LLM-as-a-Judge" node to continuously evaluate extraction accuracy.

PydanticGemini 3 FlashGemini 2.5 Pro

Podcast Generation Engine

Developed an automated B2B TTS pipeline transforming executive briefings into multi-speaker podcasts. Features contextual abbreviation expansion and artifact-free audio cloning.

Gemini 2.5 FlashVibe VoiceDemucs