Carmona / Cavite / PH
ALEF/ONE / one-man software company / AI + full-stack
ALEF/ONE is my one-man software company. I’m Alfredo Dela Peña III, an AI engineer and full-stack software engineer in the Philippines. I build OCR, fraud checks, LLM extraction, evals, dashboards, infra, agent memory, and the boring glue that keeps production systems alive.
OCR
FRAUD
EVALS
AGENTS
FULL-STACK
MEMORY
INFRA
SHIP
At Boost I work on AI Parser Studio: OCR, multi-provider LLM extraction, field dictionaries, fraud scoring, monitoring, and failover around documents that feed onboarding decisions.
Memorantado keeps local memory practical. Lode maps the codebase so I stop digging for where things live. Less ceremony, more working pipe.
The job is rarely one layer. I wire the loop: find the broken edge, cut dead process, verify with boring commands, and ship the smaller system that survives users.
AI Parser Studio: production FastAPI document processing, OCR, structured extraction, fraud detection, batch parsing, monitoring, and multi-provider failover.
Engineering lead for AI and full-stack software work.
The company itself: client work, products, internal tools, agent memory, evals, and deployment glue under one roof.
CurrentDental booking and intake product with tenant branding, QR check-in, public flows, and availability rules.
Client via BoostDocument parsing and fraud-detection gaps, including payslip AI-edit and inpainting cases flagged by risk teams.
AI / full-stackAI evaluation, contract and property document analysis, and full-stack product work around legal workflows.
Founding engineerAI-powered tax co-pilot work for creators and founders with Tax Score, strategy matching, CPA validation, and planning.
Pet health platformBackend performance, DynamoDB migration, admin dashboards, eHealth Cards, smart pet IDs, and pet-record workflows.
Real estate productProperty product work around listings, recommendation surfaces, and real-estate UX.
Community productReal-time chat, community tooling, influencer workflows, event tracking, and responsive management surfaces.
I made my own local-first memory because logging in just to put rows in SQLite was too much ceremony.
02 / codebase knowledge graphLocal repository knowledge graph CLI. Codebase discovery, symbol search, context-pack generation, and fast source-cited answers before falling back to broader search.
03 / terminal agentsForked terminal agent multiplexer. Multiple agents, less circus.
04 / lintingForked Plankton, ported it for Pi, and added Elixir support.
05 / loggingLogger library with an unserious name and a practical reason to exist.