Identify yourself. How should I reconfigure this portfolio for you?
About
Founder & Lead AI Engineer at Real Solutions PH, a software studio shipping production AI products for clients in the US, Europe, and Asia. Former Head of Engineering at ArnoldAI. Specializes in production RAG, multi-agent orchestration, real-time voice agents, and multi-tenant SaaS, and runs delivery end to end: discovery, specification, agent-assisted TDD builds, evaluation, and phase sign-off.
Technical Arsenal
AI & Machine Learning
- LangChain
- LangGraph
- PydanticAI
- RAG Systems
- LangSmith
- PyTorch
- OpenAI API
- Scikit-learn
- Pandas
- Voice AI (Pipecat, Deepgram, ElevenLabs)
- Milvus
- Agentic Coding (Claude Code, Codex)
Back-end Development
- Python
- FastAPI
- SQLAlchemy
- SQLModel
- Pydantic
- Node.js (Express)
- PHP
- Celery
- Dagster
- Alembic
Front-end Development
- ReactJS
- Next.js
- TypeScript
- TailwindCSS
- Ionic
- React Native
- Expo
DevOps & Cloud
- Docker
- Terraform
- Ansible
- GitHub Actions
- AWS
- Google Cloud
- Azure
- Hetzner
- Traefik
- Kubernetes
Database
- Firestore
- PostgreSQL
- Supabase
- MySQL
- Redis
- Row-Level Security (RLS)
Testing & Delivery
- Pytest
- Playwright
- Cypress
- Test-Driven Development
- Mutation Testing
- Scrumban
Experience
Real Solutions PH
Founder & Lead AI Engineer
📍 Philippines (Remote, global clients)|📅 2025-12 — Present
Founded a software studio that designs, builds, and operates production AI and SaaS products for international clients, on fixed-scope phased engagements.
- Delivered a voice-cloned AI coaching platform: real-time voice sessions (Pipecat, Deepgram STT, ElevenLabs TTS) grounded by RAG over Milvus, with Dagster ingestion pipelines.
- Built a multi-tenant AI agent platform with embeddable chat widgets, an operator console, tenant-isolated knowledge bases, and a bilingual dashboard.
- Shipped a real-estate market intelligence product: Dagster data pipelines feeding a FastAPI + Next.js analytics layer on Supabase.
- Engineered features for a restaurant management platform (FastAPI, Celery/Redis, React 19) serving live operations, with Playwright end-to-end coverage.
- Authored the studio's reusable full-stack template (FastAPI, SQLModel, Next.js 16, Expo, Supabase Auth, Postgres RLS), with compliance controls and evidence built in.
- Designed an agent-orchestrated delivery system: parallel worktree build sessions, architect-gated plans, test-first implementation, independent review, and traceable acceptance evidence.
- Own the full client lifecycle: discovery, proposals and SOWs, technical design, phase sign-off, and hypercare.
Tech Stack:
- FastAPI
- SQLModel
- Next.js
- React
- TypeScript
- Expo
- Supabase
- PostgreSQL
- Milvus
- Dagster
- Pipecat
- Deepgram
- ElevenLabs
- Celery
- Redis
- Playwright
- Docker
- AWS
- GitHub Actions
- Claude Code
ArnoldAI
Head of Engineering
📍 Vienna, Austria (Remote)|📅 2024-01 — 2025-11
Leading engineering for an AI platform providing chatbot agents, agent builder, and RAG systems for highly regulated industries.
- Architected Multi-Agent Systems (MAS) using LangGraph, implementing ReAct agents and Plan-and-Execute patterns.
- Engineered high-accuracy Retrieval-Augmented Generation (RAG) pipelines for complex medical and defense protocols.
- Implemented automated evaluation pipelines using LangSmith (LLM-as-a-Judge) to audit agent interactions.
- Managed a team of 8, establishing Scrumban workflows and coding guidelines.
- Migrated infrastructure from AWS to Hetzner bare-metal VMs using Terraform and Ansible, deploying via Kubernetes.
Tech Stack:
- Next.js
- React
- FastAPI
- SQLAlchemy
- PostgreSQL
- Docker
- AWS
- Azure
- GitHub
- Git
- Github Actions
- Linux
- Beautiful Soup
- OpenAI
- Pandas
- Stripe
- Supabase
- Wordpress
- Schadcn UI
- Nginx
- Figma
- PydanticAI
- Langchain
- LiteLLM
- Traefik
- Expo
- Hetzner
CoolRiots
Software Engineer (Contract)
📍 Singapore (Remote)|📅 2025-01 — 2025-02
Architected a 'Generative UI' system that dynamically renders interactive dashboards from natural language queries.
- Designed a Generative UI system using LLM Function Calling to render adaptive React components.
- Pivoted from complex multi-agent proposal to streamlined single-agent architecture via prompt optimization.
- Replaced Server-Side Rendering (SSR) overhead with client-side strategies to optimize performance.
Tech Stack:
- FastAPI
- SQLAlchemy
- PostgreSQL
- OpenAI
- React
- PydanticAI
Selected Projects
AI-based Dynamic Portfolio
Open SourceA modern, dynamic portfolio built with Next.js, designed to uniquely tailor content presentation using AI.
- Uses Google Vertex AI (Gemini) to categorize and prioritize sections based on visitor interaction.
- Data-driven design powered by a central profile.json configuration.
- SEO and metadata control centralized in a single configuration file.
- Next.js
- React
- Vertex AI
- Gemini
- TypeScript
Dynamic UI: Generative AI Component Rendering
ConceptDesigned a 'Generative UI' Proof-of-Concept that utilizes LLM Function Calling to dynamically generate structured JSON payloads, rendering adaptive React components.
- Utilized LLM Function Calling for dynamic structured JSON generation.
- Rendered adaptive React components (charts, forms, tables) based on natural language.
- FastAPI
- OpenAI API
- Pydantic AI
- Next.js
- TypeScript
Bath and Bark
CommissionE-commerce Management System.
- Engineered a serverless backend using FastAPI, SQLModel, and Pydantic.
- Built full inventory and appointment management system with Swagger API documentation.
- FastAPI
- SQLModel
- Pydantic
- Vercel
Trashbox-ARD
Research/ProjectRobust Computer Vision Model.
- Developed solid waste image classification model robust against adversarial attacks.
- Engineered for space and time efficiency using Knowledge Distillation.
- Python
- PyTorch
- Computer Vision
- Adversarial Training
Artemis - SWMS
College ProjectSmart Waste Management System.
- Built an IoT-connected waste tracking system with Express.js and Firebase backend.
- Implemented real-time monitoring via Firestore and automated CRON tasks.
- Express.js
- Firebase
- IoT
- Firestore
Clinical-RAG
Open SourceProduction-grade, citation-backed AI system designed to bridge the 'Trust Gap' in medical information retrieval.
- Orchestration: LangGraph for stateful, multi-agent workflows.
- Vector Infrastructure: Milvus for high-concurrency similarity search.
- Deployed on AWS (ECS Fargate) with CI/CD via GitHub Actions.
- LangGraph
- Milvus
- Docker
- AWS
- GitHub Actions
- Python
- Pytest
Achievements
award
Champion
@iSITE x Stackleague Programming Competition
award
2nd Runner Up
@ISEAC Hackathon - UPOU
award
3rd Runner Up
@Kadakareer x HelpBridge: Business Challenge Hackathon
award
Presentor
@ASCENT'24 - The Annual Student Conference for Engineering and Technology
award
Resource Speaker
@Seminar: Beyond the Hype - The Real World of Generative AI and Language Models
award
Participant
@iSITE National Programming Competition