Web Developer · AI Engineer

The system behind the interface.

I'm Dave Zachary Macarayo. I engineer AI pipelines, inference layers, and edge systems; then build the interfaces that make them usable.

View Projects→Contact Me

01 · Origin System

The developer behind the system.

Computer Engineering graduate turned full-stack developer with a focus on AI integration. I build practical, intelligent web systems that work.

My engineering background gives me a systems-level perspective across the full stack. I've built production web platforms with Next.js and NestJS, designed AI microservices with FastAPI and LangGraph, and deployed edge computer vision on Raspberry Pi with Coral TPU accelerators.

Whether it's orchestrating vLLM inference with RAG pipelines, implementing SOTA TTS with voice cloning, or fine-tuning YOLO for PCB component detection. I bring the same rigorous approach to software quality and system design.

Web / Full-Stack35%

AI / ML30%

Computer Vision / Biometrics20%

IoT / Embedded15%

⤢ view photo

↻ flip

$ profile --verbose

nameDave Zachary Macarayo

roleWeb Developer · AI Engineer · IoT · SQA

degreeBSc Computer Engineering

focusLLMs · RAG · TTS · Vision · Edge AI

stackNext.js, NestJS, Laravel, FastAPI, Python, Docker

statusOpen to opportunities ✦

$ card --business

name=Dave Zachary Macarayo

// engr. · Full-Stack Developer · AI Engineer

Let's build something together.

available for work

↺

✉ mdavezachary@gmail.com·DebZakari·print(name)

02 · Skill Systems

Technical capabilities.

Eight domains: web engineering, AI and ML, LLM systems, computer vision, biometrics, speech, databases, and infrastructure.

⬡

Web Development

Full-stack web engineering across modern JS and PHP ecosystems.

Next.jsNestJSLaravelCodeIgniterReactTailwind CSSAlpine.jsPHPshadcn/ui

◈

AI / ML Engineering

Model training, inference optimization, and production ML service architecture.

PythonFastAPILangGraphPyTorchTensorFlowHugging FaceComfyUIvLLMOllamaONNX Runtime

◎

LLM & RAG Systems

Retrieval-augmented generation with vector databases and semantic search.

RAG PipelinespgvectorNeo4jEmbeddingsQwen ModelsSemantic Search

SPECIALIZED DOMAINS

▣

DevOps & Infra

DockerTurborepopnpmNode.jsBullMQCloudflare R2MinIORedis

▧

Databases & ORM

PostgreSQLpgvectorNeo4jMySQLDrizzle ORMEloquent ORMRedis

○

Computer Vision

YOLOUNetTFLiteOpenCVRaspberry PiCoral Edge TPUTkinter

◉

Speech & Audio AI

Kokoro TTSFish Speechfaster-whisperVoice CloningAudio Pipelines

◇

Biometrics

Face RecognitionArcFaceRetinaFaceIris DetectionIris Segmentation

03 · Constellation Gallery

Project systems.

Web applications and AI integrations. Each one a distinct system solving a real problem.

IN PROGRESSAI

Solo Developer / Founder

NovelVerse

Turns static reading into a living ecosystem. Nova, the AI reading companion, uses Agentic Hybrid RAG (Knowledge Graph + semantic search) to let readers interrogate story lore, with Spoiler Guard built in. Authors get Git-like branch and PR workflows for collaborative editing and translations. Backed by a FastAPI microservice with LangGraph orchestration, vLLM inference, and a narration engine that voices stories aloud with expression inferred live from context, not scripted at authoring time.

Next.js 16NestJS 11FastAPILangGraphpgvectorNeo4jvLLMTurborepo

→Spoiler-aware AI reading companion + Git-like collaborative authoring

Full-Stack Developer

Enrollment V2

The original enrollment system was a single-page form: so long that users lost their place and gave up. The redesign breaks it into guided multi-step stages with progress saved to a generated code, no account required. Applicants leave and resume without starting over. Deployed in production for an academic institution and well-received.

Laravel PHPMySQLBladeTailwind CSSJavaScriptSMTP

→Multi-step enrollment with session-free progress saving; zero drop-off from form length

ML / Backend Engineer

Iris Biometric Pipeline

Full lifecycle of a production biometric system: trained the iris detection (YOLO) and segmentation (UNet) models from scratch, then built the FastAPI microservice that serves the pipeline into an existing Laravel web application. One project, full stack, from dataset to deployed API.

PythonPyTorchYOLOUNetFastAPIDocker

→Iris pipeline trained and served to production; Laravel integration live

IoT / CV Engineer

PCB Vision System

Client-commissioned edge AI system built end-to-end: trained a custom YOLOv8 model on their PCB dataset, designed the full AI-to-UI interaction flow, and built a lightweight Tkinter interface for a Raspberry Pi 4. Coral Edge TPU acceleration delivers real-time inference at 2-3x the speed of ncnn. Boots with a branded splash screen and auto-starts the detection pipeline.

PythonYOLOv8TkinterRaspberry Pi 4Coral Edge TPUONNX

→Real-time PCB inspection on edge hardware; 2-3x inference speedup via Coral TPU

04 · Mission Logs

Technical notes.

Short writeups on systems, experiments, and engineering decisions.

$tail -n 3 logs/mission.log4 entries →

›01.log|NovelVerse

2025·04

Building a Multi-Modal AI Reading Platform

Architecting a full-stack web novel platform with LangGraph agents, vLLM inference, RAG semantic search via pgvector, and SOTA TTS with voice cloning.

›02.log|Speech AI

2025·02

Kokoro + Fish Speech: Local TTS Pipeline

Integrating ONNX Runtime TTS with Fish Speech S2 Pro for voice cloning. Latency optimization, audio caching via R2, and Redis-backed semantic caching.

›03.log|Computer Vision

2024·12