Tharanitharan K.

Key Skills

Software

Other

Top Subject Matter

RAG document intelligence over PDFs

LLM orchestration and gateway for multi-provider inference

Top Data Types

Document

Text

Top Task Types

Question Answering

Function Calling

Freelancer Overview

KnowledgeAI — RAG Document Intelligence Platform (Self-hosted). Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other. Education includes Bachelor of Engineering, Karpagam College of Engineering (2025). AI-training focus includes data types such as Document, Computer Code, and Programming and labeling workflows including Question Answering and Function Calling.

Labeling Experience

AI Gateway Platform — Multi-Provider LLM Orchestration Engine (Node.js/Express/PostgreSQL/Docker)

OtherFunction Calling

Engineered an OpenAI-compatible AI gateway that orchestrates inference requests across multiple LLM providers (Ollama, vLLM, Gemini) via a provider abstraction layer. Added operational controls including dual authentication (JWT and hashed API keys), per-key IP whitelisting, plan-based rate limiting, and atomic per-token credit billing. Implemented usage analytics, conversation history with context replay, and provider health monitoring with automatic database synchronization. • Provider abstraction layer for routing inference requests • JWT + API key authentication and IP whitelisting • Per-token credit billing and request/token rate limiting • Usage analytics, audit trail, and health monitoring with DB sync

2026 - Present

KnowledgeAI — RAG Document Intelligence Platform (Self-hosted)

OtherDocumentQuestion Answering

Built a production RAG platform that enables semantic search and context-aware Q&A over uploaded PDFs using a self-hosted document intelligence pipeline. Extracted and chunked PDF text, created embeddings, and stored them in a vector database, then supported a streaming chat UI with source attribution and similarity scoring. Implemented retrieval + generation behavior so answers are grounded in the most relevant document passages. • Ingestion pipeline for PDF extraction, chunking, and embedding • Vector search using Qdrant for context retrieval • LLM-based QA and chat with context-aware responses • Source attribution and similarity score display

2025 - 2025

Education

Work History