Amario Fausta H.

I'm an AI Engineer with nearly 3 years of experience building and deploying AI systems end to end — from data collection and training through to production. My work covers two main areas — Computer Vision and LLMs / AI Agents — plus everything needed to make them actually work in the real world: data preparation and annotation, deployment and optimization, monitoring dashboards, and workflow automation around them. I currently work as an AI Engineer at an Indonesian telecom company. I mostly work with people who have an AI idea — or a system that's half-working — and need someone to get it running reliably. Not a demo. Something that stays up. Here's the detail: Computer Vision — object detection (YOLO, PyTorch), OCR, segmentation, keypoint detection, and image search with embeddings. I build these end to end: data scraping and preprocessing, training, evaluation, then deployment. My models run live on a 3,000+ camera network — face recognition at 1,000+ identities with sub-50ms latency, license plate recognition at 93% in production, waste segmentation at 86% mAP, flood and fire detection on live CCTV. I also handle the messy real-world parts: live video-stream processing, detecting dropped or frozen streams, and fixing glitches, flickering detections, and unstable tracking. LLMs & AI Agents — RAG systems, multi-agent architectures with LangGraph and supervisor routing, and tool integration (Search, Calendar, Email). Running on Ollama, vLLM, or HuggingFace, with FastAPI streaming and Redis/PostgreSQL for persistent multi-turn sessions. I've built ISP customer support agents and a WhatsApp assistant that handle real user traffic. Data prep & annotation — automated annotation pipelines that cut manual labeling effort by 50%, plus preprocessing and dataset curation so your models train on something clean. Production & optimization — this is where most AI projects stall. I've cut inference latency from 72ms to 9.9ms with TensorRT (NVIDIA DLI certified), and reduced infrastructure costs 40–60% through microservice architecture and pipeline design. FastAPI, OpenVINO, ONNX, RabbitMQ, Docker — edge or server. Monitoring & dashboards — so you can see how your system is actually performing, with alerts when something needs attention. I run this at scale across 300 servers. Document & workflow automation — pulling data out of PDFs, invoices, and scanned documents into clean CSV/Excel/Sheets, and connecting repetitive steps (email, files, reports) into one flow that runs on its own. I keep things clear and maintainable: clean code, written documentation, and a handover call so your team can keep it running after we're done. I won't disappear after delivery. I'm available around 20 hours a week. If you're stuck on a CV, LLM, or automation problem, tell me what you're trying to do — I'll be honest about whether I'm a good fit and how I'd approach it.