adrian.fahri affandi/portfolio'26
AI Engineer · Backend Infrastructure

Adrian.
builds quiet infrastructure

Software Engineer specializing in AI integration and backend infrastructure. I take ML models from experimentation to production — edge AI systems, RAG pipelines, and the microservices backbone that keeps them running reliably in Go and Python.

shipped
5+ systems
edge AI, backend, IoT
stack
Go · Python
+ PyTorch, WebRTC
currently
AI Engineer
at Synapsis · Oct 2025
education
B.Eng IST
ITB · graduated 2025
§ 01

Selected work

Two systems I shipped end-to-end. Click to expand.
01
2025

Seismic Risk

CV pipeline for building typology classification — ITB thesis
Solo engineer, end-to-end

Microservices backend for geospatial seismic risk assessment. YOLOv11 models classify building typologies; an integrated annotation platform replaced third-party tools and cut manual labeling effort by 40%+.

  • 01Microservices backend with FastAPI + PostgreSQL/PostGIS for geospatial data management
  • 02Trained and deployed YOLOv11-based CV models for building typology classification
  • 03Custom MLOps workflow with model versioning and rollback support for reproducible experiments
  • 04Integrated annotation platform with pre-annotation that reduced manual labeling by 40%+
  • 05AI predictions fed into a seismic risk simulation covering 10,000+ buildings region-wide
PythonFastAPIPostgreSQLPostGISYOLOv11PyTorchDocker
[SEISMIC RISK · diagram]
░ system architecture ░
buildings
10k+
labeling ↓
40%+
model
YOLOv11
§ 02

Experience

Where I've been shipping
total experience
3+ yrs
including internship + research
roles
3
intern → junior eng
production systems
2 shipped
owned end-to-end
team size today
5 engineers
ML + backend
Full-time · 8 mo

AI Engineer

Synapsis·Indonesia
current

Architected and deployed production AI systems — from edge-to-cloud face recognition for enterprise attendance to a multi-agent LLM ecosystem for internal support.

  • Deployed Edge-to-Cloud Multi-Face Recognition with Raspberry Pi 5 + Hailo AI accelerators for 750+ employee attendance
  • Increased video processing throughput 150% (8 FPS → 20 FPS) by decoupling AI inference via multi-threading
  • Built production-grade multi-agent LLM ecosystem with a centralized master agent for dynamic query routing
  • Developed advanced RAG pipeline with Perplexity-style inline citations, boosting document traceability
PythonGoRaspberry PiHailo AIRAGLangGraphFastAPI
§ 03

Stack

Tools I reach for first
Languages
  • Go
  • Python
  • TypeScript
  • Java
  • SQL
  • C++
AI / ML
  • PyTorch
  • YOLOv11
  • LangGraph
  • RAG / LLM
  • Hailo AI
  • OpenCV
Backend
  • FastAPI
  • NestJS
  • gRPC
  • PostgreSQL
  • MySQL
  • Redis
Infra
  • Docker
  • Kubernetes
  • CI/CD
  • Raspberry Pi
  • WebRTC (LiveKit)
§ 04

About

The short version
I graduated from ITB in 2025 with a B.Eng in Information System & Technology. Over the past year I've been shipping production AI systems at Synapsis — from edge-deployed face recognition on Raspberry Pi 5 + Hailo to multi-agent LLM pipelines. I'm most useful at the intersection of ML and backend: designing the services that turn a trained model into something reliable, low-latency, and on-call-able.
education
B.Eng Information System & Technology
Institut Teknologi Bandung (ITB) · GPA 3.44
previously
Backend Dev → AI Engineer
Synapsis & Telkom Indonesia · 2024–present
currently
AI Engineer
Edge AI + LLM systems, Synapsis
looking for
AI / Backend roles
Edge AI · LLM infra · backend systems
§ 05

Writing

Notes, posts, postmortems
2026.04 Building a Multi-Agent LLM Ecosystem for Internal Support post 2026.02 Edge AI with Hailo: Getting 20 FPS Face Detection on Pi 5 post 2025.11 YOLOv11 for Building Typology — Lessons from a Thesis notes 2025.08 What I Learned Moving from Backend Intern to AI Engineer notes