Jerome Agapay

About

Full-stack engineer with 8 years of experience, now focused on building AI/RAG systems. During a two-year gap, I built personal projects to broaden my depth in software engineering and AI. Actively looking for my next role — open to interesting problems.

I code for a living 👨🏻‍💻

I read whatever pulls my attention 👀

I surf occasionally

I run on Tuesdays and Saturdays (most weeks)

I travel solo (yes, on purpose)

Connect with me: jerome.arceo.agapay@gmail.com, Resume, LinkedIn

Technical Skills

  • Programming Languages: Javascript, Typescript, Python, PHP 7.x, SQL
  • Frontend Technologies: ReactJS, NextJS, Redux, React Native, Vite, Astro, Tailwind CSS, Radix UI, TanStack Table, Ant Design, jQuery
  • Backend Technologies: NodeJS, Express.js, FastAPI, Serverless Framework, Laravel, Redis, NextAuth
  • Databases: PostgreSQL, MySQL, MariaDB, Aurora Serverless v2, Amazon RDS, SQLite, Prisma, Drizzle ORM, Liquibase
  • AI/RAG: LlamaIndex, LangChain, Ollama, RAG pipelines, vector search (Qdrant, ChromaDB), hybrid retrieval & reranking, embeddings, RAGAS evals, LLM-as-judge evals
  • Cloud/AWS: Lambda, API Gateway, RDS, Aurora Serverless v2, S3, S3 Glacier, CloudFront, Cognito, CloudWatch, SQS, EventBridge, SES, EC2, VPC, ECS Fargate, AWS Batch, ECR, Secrets Manager, Route 53, ACM, IAM, Amplify, CDK, CloudFormation
  • AI Tools: Claude Code, Anthropic API, AI-assisted development, LLM integration
  • Development Tools: Git, Docker, Docker Compose, CI/CD pipelines, Vitest, Agile/Scrum methodologies
  • Operating Systems: Linux(Ubuntu), macOS, Windows

Certifications

AWS Certified Solutions Architect - Associate (SAA-C03)

Oct 2025

Work

HealthyLife Company Pty Ltd.

Senior Software Engineer

July 2022 - May 2024

  • E-commerce and digital health platform.
  • Stabilized storefront by removing recurring UI defects and consolidating patterns across pages.
  • Modernized legacy jQuery modules into reusable components and simplified CSS with Tailwind.
  • Improved mobile behavior and layout consistency across high-traffic views.
  • Worked within legacy constraints while raising code quality through small, safe refactors.
  • Technologies used: HTML, CSS, jQuery, SASS, Tailwind CSS, PHP, Laravel

PayMaya Philippines Inc.

Senior Software Engineer

May 2021 - July 2022

  • Business management platform for payment, banking, and loan solutions.
  • Designed file-based disbursement workflow using S3 -> SQS -> Lambda -> RDS with DLQs and idempotency.
  • Built CRUD services with Serverless Framework for React front-end and Node.js back-end.
  • Ran targeted load tests to validate batch processing behavior and tune concurrency.
  • Collaborated in reviews and sprint planning with engineers, QA, and product.
  • Technologies used: React 17, Redux, Node.js, AWS Lambda, SQS, RDS, S3, Serverless Framework

Botbros AI, Rocket Labs Consulting Inc.

Jun 2017 - Jan 2021

Airship Logistics Management System
Tech Lead
  • Led full-stack delivery for bookings, agent management, and cash-flow operations.
  • Drove Agile cadence and code reviews to improve consistency and maintainability.
  • Researched and applied AWS patterns for scale and reliability.
  • Recognized Top 3 in the IdeaSpace Acceleration Program.
  • Technologies used: React, Redux, Node.js, Serverless, AWS (Lambda, API Gateway, RDS, Cognito, CloudWatch, CloudFront, S3).
PeopleServe — Human Resource Information System
Software Engineer
  • Scoped with clients to clarify requirements and realistic timelines.
  • Mentored interns and junior developers to raise delivery quality.
  • Provided production support and issue resolution post-launch.
  • Technologies used: Laravel 5.2, React, Redux, MariaDB, MySQL 5.6.
Spectrum Investments - In-house API
Software Engineer
  • Created database schema.
  • Written documentation for the applications custom http requests and responses.
  • Collaborated with freelance developers.
  • Technologies used: PHP Slim 3
Entrego - Booking Website
Software Engineer
  • Coded their Homepage, About, Contact, Career and Booking pages.
  • Integrated Entrego's API with the login and booking pages.
  • Cooperated with Zalora team with regards with some changes in their booking website.
  • Suggested and implemented changes in the UI.
  • Created reusable components using React.
  • Technologies used: HTML, CSS, React JS, Redux
Chatbots - In-house projects
Software Engineer
  • Research and developed facebook messenger and slack bots.
  • Technologies used: Botkit JS, Chatfuel, Node JS
HonestBee - Booking Website
Software Engineer
  • Created BotBros AI website using HTML, CSS and JS.
  • Modified the implemented PHP back-end.
  • Technologies used: vanilla HTML, CSS and JS, PHP 7.1.
BotBros AI - Website, Junior Full Stack Engineer
Software Engineer
  • Created BotBros AI website using Ghost CMS, HTML, CSS and JS.
  • Technologies used: Ghost CMS, vanilla HTML, CSS and JS.

Ideyatech Inc.

QA Intern

Oct 2016 - Jan 2017

  • Authored test cases that expanded coverage of core web flows.
  • Performed manual functional testing and documented issues for fast fixes.

Projects

PH Law RAG

Local-first RAG assistant over Philippine law with hybrid search, reranking, and grounded citations

Apr 2026 - Present

  • Link to github repository
  • Built a local-first Retrieval-Augmented Generation assistant over 45 Philippine law primary sources, running generation and embeddings on local models with no hosted inference API.
  • Designed a hybrid retrieval pipeline combining dense vector search (Qdrant) with sparse BM25, followed by Qwen3 cross-encoder reranking for higher answer relevance.
  • Implemented incremental document sync via content hashing so re-runs skip unchanged statutes, with grounded generation that cites article/section numbers and abstains when the corpus can't support an answer.
  • Built an eval harness on RAGAS with an Anthropic LLM-as-judge, plus deterministic retrieval comparison runs to measure the impact of each retrieval change.
  • Deployed to AWS with CDK across persistent and application stacks: ECS services, ECR images, Cloud Map service discovery, Secrets Manager, ACM, and Route 53.
  • Documented 23 architecture decision records and backed the system with 58 test modules plus structured tracing and logging.
  • Technologies used: Python, LlamaIndex, Ollama, Qdrant, BM25, cross-encoder reranking, RAGAS, Anthropic API, FastAPI, React, Vite, Tailwind CSS, SQLite, Docker, AWS CDK, ECS

AWS & Amazon Bedrock Docs RAG Lab

Local-first RAG system over AWS and Amazon Bedrock documentation with incremental sync and grounded citations

Apr 2026

  • Link to github repository
  • Built a local-first RAG system over curated AWS and Amazon Bedrock documentation that answers questions with a local LLM and cites the source docs for verification.
  • Implemented a document ingestion pipeline that fetches an allowlist of AWS and Amazon Bedrock pages, normalizes HTML to clean text, and detects changes via content hashing to only reprocess what changed.
  • Curated Bedrock documentation covering model inference, Knowledge Bases, Guardrails, Agents, Flows, batch inference, and evaluations as the primary RAG corpus.
  • Chunked and embedded documents locally with sentence-transformers, stored vectors in ChromaDB, and tracked document metadata and versions in SQLite.
  • Exposed the pipeline through both a Typer CLI and a FastAPI service, with an eval harness scoring answers against a question dataset.
  • Technologies used: Python, Ollama, sentence-transformers, ChromaDB, FastAPI, Typer, SQLite, BeautifulSoup, Amazon Bedrock docs

PSilo (Personal Silo)

Self-hosted image storage platform built as an alternative to commercial cloud storage solutions

Feb 2026 - Apr 2026

  • Link to github repository
  • Architected and built a full-stack personal cloud storage app using AWS CDK for infrastructure-as-code, provisioning Cognito, API Gateway, Lambda, SQS, EventBridge, Aurora Serverless v2, CloudFront, SES, and S3 with Glacier Flexible Retrieval.
  • Implemented cost-optimized cold storage by transitioning originals to S3 Glacier via lifecycle rules, with a storage dashboard showing per-storage-class cost and retrieval cost estimates.
  • Built the Glacier retrieval flow end to end: restore tier selection (Expedited/Standard/Bulk), EventBridge-driven completion handling, SES notification, batch tracking, and album downloads zipped by an ECS Fargate pipeline.
  • Added video support with an AWS Batch on Fargate Spot pipeline running FFmpeg for thumbnail covers and hover previews, and pHash perceptual duplicate detection at upload time.
  • Served thumbnails and previews through CloudFront with signed URLs and 24-hour edge caching to cut origin requests and egress cost.
  • Designed monorepo structure separating frontend (Next.js), infrastructure (per-domain CDK constructs), and Lambda services, with async metadata processing via SQS and a dead letter queue.
  • Integrated Drizzle ORM with Aurora Serverless v2 via the Data API (no VPC/NAT) with migrations for type-safe SQL and version-controlled schema management.
  • Technologies used: Next.js, TypeScript, Node.js, AWS CDK, Lambda, API Gateway, Aurora Serverless v2, S3 Glacier, CloudFront, Cognito, SQS, EventBridge, SES, AWS Batch, ECS Fargate, ECR, Drizzle ORM

ATC Monitoring System

Custom client website replacing Excel system workflows

May 2025 - Present

  • Link to github repository
  • Eliminated duplicate records and manual errors by moving Excel-based operations into a typed web app covering containers, inventory, auctions, bidders, payments, expenses, payroll, and petty cash.
  • Modeled the business domain across 28 relational tables in Prisma against MySQL, with inventory history and activity audit logs for traceability.
  • Built a full payroll module (pay periods, earnings, deductions, employment events) and financial and owner-profit reports exported to PDF and Excel.
  • Migrated the deployment from Vercel to self-managed AWS, authoring a CloudFormation template that provisions a VPC with public and private subnets, NAT gateway, security groups, EC2 with an IAM instance profile, RDS MySQL, and S3, with CloudFront in front of the app.
  • Added authenticated S3 uploads for container reports using presigned URLs and the EC2 instance profile, avoiding long-lived credentials.
  • Used an internal RPC approach instead of REST since no public API was needed, with clean architecture (use cases and repositories behind dependency injection) and 52 test suites.
  • Technologies used: Next.js, TypeScript, Prisma, MySQL, NextAuth, Zod, TanStack Table, Recharts, RPC, AWS EC2, VPC, RDS, S3, CloudFront, CloudFormation