Gilang Hidayatullah
I build grounded, evaluated AI systems — most recently Gizinesia, an Indonesian nutrition calculator with zero hallucinated numerical claims across 30 eval scenarios.
Case Studies
Gizinesia — AI Nutrition Calculator
Problem
Indonesian nutrition students and dietitians lack an accurate, TKPI-grounded nutrition calculator; generic AI tools hallucinate nutritional values.
Solution
Next.js + Supabase + Vercel app. Function-calling architecture: the LLM handles only natural language, while deterministic tools perform every calculation against TKPI 2020 / AKG 2019 data. Multi-model support via OpenRouter.
Pepper Humanoid Robot — Cloud-Offloaded HRI
Problem
Resource-constrained humanoid robots can't run modern CV/LLM workloads on-device.
Solution
Cost-effective cloud-offloading architecture — face recognition and registration flow on the robot, heavy inference on GCP (Cloud Run, Vertex AI), Redis as temporary store with keyframe optimization for efficiency.
Highlights
Real-time device → API → response interaction loop; production deployment patterns on GCP.
Evolipia Radar — AI Research Signal Aggregator
Problem
AI research output is overwhelming; engineers need a verified, prioritized signal feed.
Solution
Go backend aggregating and ranking AI research signals (Engineering Verified Overview of Latest Insights, Priorities, Impact & Analytics).
More Projects
supplychain-ml-forecasting-optimizer
End-to-end retail supply chain analytics demo: generate dummy data, engineer features, train weekly demand forecast model, and optimize replenishment under budget using linear programming. Exposed via FastAPI service.
Websocket-AI
Go application providing WebSocket interface for communicating with Google Vertex AI, supporting real-time text and audio input with Text-to-Speech (TTS) responses.
ai-task-analyst
Intelligent AI-powered web application for analyzing and managing project/task data, featuring advanced function calling, data analysis, and AI chat that remembers previous conversations and provides actionable insights.
ML Demonstrations
Experience real-time machine learning powered by TensorFlow.js and face-api.js. All processing happens in your browser for maximum privacy.
Hand Gesture Game
Play rock-paper-scissors using hand gestures detected in real-time. Uses TensorFlow.js Handpose model for accurate gesture recognition.
Face Expression Detector
Detect emotions from facial expressions in real-time. Analyzes 7 different emotions with confidence scores using face-api.js.
Object Detection Game
Find objects using your camera in this interactive game. Uses COCO-SSD model to detect 80+ different object types in real-time.
Privacy First: All ML processing happens locally in your browser. No video or images are ever sent to external servers.
Skills Overview
Languages
AI / Machine Learning
Cloud & Infra
Backend & APIs
About Me
Hello! I'm an AI & Machine Learning Engineer specializing in production-oriented AI on GCP, robotics, and backend systems. I enjoy transforming complex requirements into reliable solutions and recently shipped Gizinesia, a live nutrition calculator, as a solo founder.