About
Software Engineer with 3+ years of experience building and deploying production-grade AI-powered systems end-to-end. Proficient in Python and TypeScript, with hands-on experience integrating LLMs and building scalable backend services using Flask, Django, and Kubernetes (GKE). Proven ability to own full systems—from architecture design to cloud deployment—in fast-moving, ambiguous environments. Passionate about leveraging AI to create products that make a real impact.
Currently
- Architecting a multi-stage AI ensemble for smartphone grading on GKE
- Tuning HPA + observability with VictoriaMetrics & Grafana
- Building Axiom Alpha — institutional-grade equity analysis at axiomalpha.online
- Exploring agentic LLM workflows for engineering productivity
Work Experience
Tech Stack
languages
backend
ai / ml
llm & rag
mlops & cloud
databases
tools
Check out my latest work
I've led a range of AI initiatives—from rapid proof-of-concept models to production-scale ML systems like multi-stage computer-vision pipelines, RAG-based assistants, and LLM-powered automation. Here are a few highlights.
Axiom Alpha
Institutional-grade equity analysis platform with multi-factor scores (Growth, Quality, Value, Momentum, Risk Shield), portfolio backtesting, market heatmap, and a curated screener spanning US and Indonesian markets. Architected the Django backend, designed the frontend, and run end-to-end cloud deployment.
Crossed-Eye (Strabismus) Detection
Computer vision system that detects strabismus (crossed eyes / mata juling) from facial images. Uses landmark detection and deep-learning classification to flag misaligned gaze axes, intended as a pre-screening aid.
AI-Recruitment
NLP + RAG recruitment pipeline using OpenAI embeddings, Pinecone vector store, and a FastAPI inference service. Designed to accelerate candidate screening and track GPT cost analytics.
YOLOv11 Detection API
FastAPI + Docker service for serving YOLOv11l. Users upload images, receive detection results, and download annotated images via a static file server.
Person Detection (Faster R-CNN + COCO)
Faster R-CNN with a GoogleNet backbone for person detection on COCO (3,000 train / 500 test) using FiftyOne. Achieved mAP 0.245.
Cityscapes UNet Segmentation
UNet implementation on the Cityscapes dataset—Adam optimizer, LR 1e-3, 100 epochs, resize-256 augmentation. Reached Dice Coefficient 0.912.
Certifications & Achievements
Building Systems with the ChatGPT API
DeepLearning.AI
Prompt Engineering with Llama 2 & 3
DeepLearning.AI
Supervised Machine Learning: Regression and Classification
DeepLearning.AI / Coursera
Full Stack Data Science Academy
Indonesia AI
Database Programmer
BNSP (Indonesian National Certification)
AI Career Bootcamp for Computer Vision — 2nd Place
Indonesia AI Bootcamp Batch 1