Philippines

Aries Lacson

Senior AI / LLM Backend Engineer

I build reliable RAG platforms, recommendation systems, and production backend services that turn AI capabilities into measurable business outcomes.

View selected workContact me
RETRIEVAL_GRAPHRAG · SEARCH · BACKEND

Outcomes from professional engagements — not universal guarantees

10+

Years of engineering experience

Across backend, distributed systems, and AI/LLM platforms.

35%

Reduction in manual analyst review

Achieved by automating retrieval and validation workflows on a production AI platform.

26%

Reduction in average LLM cost

Delivered through prompt, retrieval, and model-routing optimization.

90%+

High-risk regressions caught before release

Through structured validation and observability built into the release pipeline.

About

Backend rigor, applied to AI systems

Aries combines deep backend engineering in Java Spring Boot and Python with hands-on production experience building retrieval-augmented generation (RAG) systems, semantic retrieval, recommendation engines, structured output validation, observability, and distributed systems that operate reliably at scale.

Before backend engineering, he worked as an architectural designer — a background that still shapes how he approaches software today: structured systems thinking, careful visual and information architecture, and disciplined documentation of complex designs.

Expertise

Where I focus

Core areas of depth, distilled from a decade of backend and AI systems work.

AI & LLM Systems

Designing and operating retrieval-augmented generation pipelines and LLM-powered services in production.

RAG pipelinesPrompt & context engineeringSemantic retrievalModel routing & cost optimizationStructured output validation

Backend Architecture

Building resilient backend services and APIs with Java Spring Boot and Python.

Java Spring BootPython servicesREST & event-driven APIsDomain-driven designService reliability

Search & Recommendation

Building discovery and ranking systems that surface relevant results at scale.

Recommendation enginesSemantic & hybrid searchRanking & relevance tuningPersonalization

Data & Vector Infrastructure

Designing data pipelines and vector infrastructure that keep retrieval accurate and fast.

Vector databasesEmbedding pipelinesETL & data pipelinesKnowledge graphs

Distributed Systems

Architecting systems that stay correct and available under real-world scale and failure.

MicroservicesEvent streamingCaching & consistencyFault tolerance

Cloud, DevOps & Reliability

Operating cloud infrastructure with observability and reliability as first-class concerns.

AWS & AzureCI/CD pipelinesObservability & monitoringInfrastructure as code

Selected work

Case studies

A closer look at representative engineering work, described at a level appropriate for public sharing.

AI Knowledge Graph and RAG Enrichment Platform

01
Problem
Large volumes of crawled web content required substantial analyst review before becoming trusted, reusable structured knowledge.
Approach
Built RAG enrichment services using metadata-aware chunking, hybrid retrieval, reranking, source grounding, embeddings, and structured JSON validation.
Outcome
Reduced manual analyst review by 35% and improved extraction consistency by 18% across selected high-volume datasets.
PythonFastAPILangChainLlamaIndexPostgreSQLRedisVector Search

Recommendation and Ecommerce Discovery Backend

02
Problem
Catalog discovery produced excessive zero-result searches and slow response paths across high-traffic ecommerce journeys.
Approach
Rebuilt search, filtering, ranking, caching, chatbot, and recommendation workflows using Node.js, TypeScript, Python, PostgreSQL, Redis, and React.
Outcome
Increased product findability by 17%, reduced zero-result searches by 21%, and reduced average response time by 25%.
Node.jsTypeScriptPythonPostgreSQLRedisReactRecommendation Systems

Java Microservice Platform for Finance and Telecom

03
Problem
Manual cross-system data handoffs and slow reporting workflows reduced operational data freshness and responsiveness.
Approach
Developed event-driven Spring Boot services with REST APIs, PostgreSQL, Redis, messaging, RBAC, audit logging, retries, and idempotent processing.
Outcome
Reduced manual data handoff by 40%, brought important reports from several minutes to under 20 seconds, and reduced repeat production issues by 38%.
JavaSpring BootPostgreSQLRedisREST APIsEvent-Driven ArchitectureRBAC

Retail POS Analytics and Operations Platform

04
Problem
Retail teams relied on manual reconciliation and difficult-to-maintain modules across inventory, sales, transactions, and reporting.
Approach
Modernized Java and SQL modules into clearer service, repository, and validation layers, with administrative dashboards and REST integrations.
Outcome
Reduced manual reconciliation effort by 30% and post-release fixes by 24% in frequently updated components.
JavaSQLJavaScriptREST APIsPOS SystemsReporting

Experience

Career timeline

Senior AI / LLM Backend Engineer

Jun 2022 — Mar 2026

Diffbot

  • Architected production RAG enrichment services for crawled web content (Python, FastAPI, LangChain, LlamaIndex, vector search) with hybrid retrieval and structured validation — reducing manual analyst review by 35%.
  • Built retry-safe async LLM workers with caching and token-efficient context preparation, cutting average LLM cost per record by 26% and p95 latency by 22%, backed by golden-dataset evaluation that caught 90%+ of high-risk regressions before release.

Senior Backend & AI Systems Engineer

Sep 2020 — May 2022

Iflexion

  • Rebuilt ecommerce search, catalog discovery, and recommendation modules (Node.js, TypeScript, Python, PostgreSQL, Redis, React) — increasing product findability by 17% and cutting zero-result searches by 21% and response time by 25%.
  • Integrated NLP-assisted chatbot and recommendation flows that deflected 28% of repetitive product inquiries from support, while mentoring three engineers through code reviews, debugging, and design reviews.

Java Backend Engineer

Jun 2017 — Jul 2020

Software Mind

  • Built event-driven Java Spring Boot microservices and synchronization services for finance, telecom, and enterprise platforms, reducing manual data handoff by 40% and cutting report-generation time from minutes to under 20 seconds.
  • Hardened authentication, authorization, RBAC, and audit logging, and added retry handling, idempotent processing, and structured logging and alerts — reducing repeat production issues by 38%.

Full Stack Java Developer

Aug 2014 — Apr 2017

Exist Software Labs

  • Modernized Java-based retail and POS modules — inventory, sales, branch operations, transaction processing, and reporting — reducing manual reconciliation effort by 30%.
  • Refactored legacy Java and SQL into clear service, repository, and validation layers, and built admin dashboards and reporting interfaces, cutting post-release fixes by 24% in frequently updated components.

Architectural Designer

Oct 2010 — May 2014

SP Moderna

  • Produced architectural drawings and documentation — floor plans, elevations, and 3D visualizations in AutoCAD — while coordinating across architects, engineers, and construction teams, building an early foundation in systems thinking, visual structure, and documentation discipline.

Certifications

Credentials

Microsoft Certified: Azure AI Engineer Associate

Microsoft

AWS Certified AI Practitioner

AWS

MongoDB Associate Developer

MongoDB

Contact

Let's talk

Open to discussing AI platforms, backend systems, or senior engineering roles. Email is the fastest way to reach me.

lacsonaries542@gmail.comGitHubLinkedIn