Ardutra A.

Key Skills

Software

No software listed

Top Subject Matter

Enterprise knowledge retrieval (CMDB) for IT support using RAG

Top Data Types

Text

Image

Top Task Types

Routing

Classification

Object Detection

Freelancer Overview

IT Infrastructure Management Intern — Production-grade RAG chatbot for CMDB automation (PT. Bank Negara Indonesia). Brings 3+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include FastAPI, Qdrant, and Groq LLM. Education includes Bachelor of Electrical Engineering, Universitas Islam Indonesia (2025). AI-training focus includes data types such as Text and Image and labeling workflows including Routing, Classification, and Object Detection.

Labeling Experience

IT Infrastructure Management Intern - PT Bank Negara Indonesia (Persero) Tbk

TextClassification

As an IT Infrastructure Management Intern, they engineered a production-grade Retrieval-Augmented Generation chatbot to automate CMDB information retrieval for NOC and L1/L2 support teams. They implemented domain-aware retrieval using zero-shot classification and built a conversational query optimization component with LLM-based query rewriting for better multi-turn understanding. They also architected a dual-path routing system and added observability features to improve reliability and troubleshooting across the RAG pipeline. • Built a FastAPI-based RAG chatbot with Qdrant and Groq LLM • Designed zero-shot classification routing across CMDB knowledge domains • Developed query rewriting for multi-turn context-dependent requests • Implemented logging, exception handling, and fixed retrieval scoring defects

2025 - 2026

IT Infrastructure Management Intern — Production-grade RAG chatbot for CMDB automation (PT. Bank Negara Indonesia)

TextRouting

Developed a production-grade Retrieval-Augmented Generation (RAG) chatbot to automate CMDB information retrieval for NOC and L1/L2 support teams. Implemented a domain-aware retrieval framework using zero-shot classification to route user queries across five CMDB knowledge domains. Built LLM-based query rewriting and a dual-path query routing approach to improve contextual understanding and operational response reliability. • Reduced reliance on manual CMDB database lookups • Fixed a vector similarity scoring defect to improve retrieval precision • Added structured multi-level logging and exception handling to improve observability and troubleshooting • Enhanced conversational multi-turn question understanding for context-dependent requests

2025 - 2026

Final Project (Computer Vision Deep Learning) - Universitas Islam Indonesia

ImageClassificationObject Detection

For a capstone final project, they developed a real-time exam cheating detection system using YOLOv8 object detection and pose estimation models. They designed and optimized a computer vision inference pipeline to support classroom deployment with a focus on balancing accuracy, efficiency, and real-time performance. They evaluated multiple model configurations under varying lighting and environment conditions to improve robustness. • Built a real-time cheating detection workflow using YOLOv8 and pose estimation • Optimized inference for 1080p video streams at ~8–9 FPS • Conducted performance evaluation and comparative model configuration testing • Reported detection and pose classification accuracy across scenarios

2024 - 2025

Education

Work History