Farhan Kamil Hermansyah@BadsFarmil
1
Events attended
1
Submissions made
Indonesia
1 year of experience
About me
I specialize in building production-ready, multi-agent AI architectures and advanced Retrieval-Augmented Generation (RAG) pipelines. My focus is on bridging raw LLM capabilities with structured, deterministic business logic. My hands-on experience includes developing several complex AI systems: GraphWeaver: Engineered an investigative data mapping system utilizing Neo4j and LangGraph to autonomously extract, map, and navigate complex entity relationships from unstructured data. Domain-Specific RAG Systems: Developed ScholarSync (an AI research assistant) and ExportAI (an intelligence system for navigating complex export regulations for the Digdaya Hackathon 2026). Both utilize dynamic RAG pipelines, integrating Vector Databases (Pinecone/ChromaDB) with LangChain. Autonomous B2B Workflows: I design systems that don't just chat, but execute tasks. I integrate these cognitive AI engines into production pipelines using n8n and FastAPI, implementing strict state management and structured JSON outputs. Additionally, my work in applied Generative AI earned me a spot as a Top 20 Finalist in the AI Talent Hub Indonesia 2026. I am highly focused on building reliable, autonomous agents with hybrid database architectures (Graph + Vector) to solve real-world industry bottlenecks.
🤓 Latest Submissions
BargainHunter_B2B And Pricing Surveillance Agent
BargainHunter is an autonomous, enterprise-grade AI agent explicitly designed for the Web Data UNLOCKED Hackathon to revolutionize Go-To-Market (GTM) Intelligence and Finance strategies. Built on a robust FastAPI backend and orchestrated by a sophisticated LangGraph multi-node architecture, BargainHunter automates the critical process of real-time market and pricing surveillance. The system executes a seamless multi-step AI workflow: First, a Scraper Node leverages Bright Data to extract targeted competitor pricing while filtering out irrelevant noise. Next, an Analyzer Node aggregates this data to identify pricing anomalies and market leaders. A Recommender Node then employs Retrieval-Augmented Generation (RAG) using a Pinecone vector database to recall historical intelligence reports, providing historical context for strategy formulation. Finally, a unique Critique Node acts as a manager, utilizing a self-reflection loop to ensure all AI-generated pricing strategies strictly adhere to enterprise business rules before finalizing the output. Featuring real-time streaming execution, users can watch the AI's step-by-step logic live via a Streamlit dashboard. Furthermore, the system includes a robust B2B alerting mechanism that instantly notifies GTM and Finance teams of high-priority market shifts, empowering data-driven, strategic decision-making.
31 May 2026
