Production-Ready AI & Software Engineer

Currently Building

ForceX AI — Indonesia's first AI-for-energy platform. 12 physics-informed AI products for nuclear, geothermal, oil & gas, and renewables. GeoForce v1.1 deployed with R²=0.997 accuracy. Explore the platform →

Overview

I am Robi Dany Riupassa, a software engineer and AI specialist building production-grade AI systems — from physics-informed energy platforms to agentic RAG chatbots and enterprise MLOps pipelines.

My work spans:

  • Agentic AI & RAG — LangGraph multi-agent orchestration, self-reflective retrieval, hybrid search, multi-provider LLM fallback chains
  • Physics-Informed ML — PINNs, CNN surrogates, GNNs for energy sector simulation (ForceX AI, 12 products)
  • MLOps & Model Deployment — Automated training pipelines, model registry, 24 production models, CI/CD with GitHub Actions
  • Cloud & Infrastructure — Google Cloud Platform (Cloud Run, Cloud SQL, Secret Manager), Docker, Kubernetes, Kafka, Terraform
  • Multimodal AI — NLP (spaCy, Transformers), Vision (CLIP, YOLOv8), Audio (Whisper), Vector Search (pgvector)
  • Published Work — Models and datasets on HuggingFace, AI testing framework on npm (@robi-atp/cli)

Flagship — ForceX AI

ForceX AI — AI-for-Energy Platform | forcex-ai.com

12 physics-informed AI products (PINNs, CNN, GNN, LSTM, RL) with LangGraph agent orchestration. GeoForce CNN surrogate deployed at R²=0.997. Model and dataset published on HuggingFace. 484 passing tests.

Enterprise Systems (Private Company)

Enterprise Agentic RAG Chatbot — LangGraph multi-agent RAG with self-reflection, hybrid search (dense + BM25 + RRF), multilingual, Langfuse observability. Production-deployed.

Media Platform with OTT Streaming — 8+ NestJS microservices, multimodal AI (NLP/Vision/Audio), Kafka event streaming. OTT deployed on Google Cloud Platform (5 Cloud Run services, Cloud SQL, Compute Engine).

Broadcast Analytics — MLOps Platform — 24 trained models (6 algorithms x 4 targets), LangChain agent with 7 tools, genetic algorithm optimizer, Prometheus/Grafana/Alertmanager monitoring stack.

AI-Powered Help & QnA System — Production RAG with hybrid retrieval, 5-provider LLM fallback chain, cross-encoder reranking, voice assistant. 75%+ cache hit rate, ≤5% hallucination rate.

Independent Projects

GeoForce — CNN Reservoir Surrogate — Standalone production model replacing TOUGH2 simulation. R²=0.997, 57K params. Published on HuggingFace.

Porto Agent — AI content marketing platform. Generates blog posts from git commits using Gemini AI + LangGraph, with human-in-the-loop approval via dashboard and Telegram.

AI Testing Framework (ATP) — Playwright-like testing for AI apps. Semantic similarity, hallucination detection. Published on npm as @robi-atp/cli.

Recommendation Systems Research — YouTube Two-Tower, Netflix Foundation, and Hybrid architectures compared across 108 hyperparameter configurations.

Browse all projects →

Professional Experience

Developer & Researcher — ForceX AI | Current

Building Indonesia's first AI-for-energy platform — 12 physics-informed AI products, GeoForce deployed, models published on HuggingFace.

Software Developer & AI Specialist — A Leading ICT Solutions Provider | Current

Enterprise AI systems for media and broadcast — agentic RAG chatbots, MLOps pipelines, multimodal AI, cloud-native deployment on GCP.

Training Instructor — Government Transformation Academy, BPPTIK KOMINFO

Delivered Fundamentals of Data Science training for government professionals.

View full experience →

Education

Doctoral Research in Physics — Bandung Institute of Technology, Indonesia Passive safety systems for Generation IV Molten Salt Reactors, combining CFD simulation with deep learning. ITB Digital Library →

Master of Science (MSc) in Physics — Bandung Institute of Technology, Indonesia

View education details →

Publications

Published research in physics, computational science, and AI. Google Scholar | Scopus

View publications →

Contact