Welcome to the GitHub Site of Cruz Macias

Oracle Integration Cloud (OIC) Engineer & Enterprise Systems Architect (SCM/FIN/HCM) — building, testing, and maintaining enterprise integrations and technical extensions supporting Oracle Fusion Cloud ERP under Kanban AGILE methodology and change control management. M.S. Computer Science — Cybersecurity & Digital Forensics, EKU 2023. B.S. Mathematical Sciences, NIU 2016.

Link to Curriculum Vitae — Curriculum Vitae

Link to Music Portfolio — Music Portfolio

Notable Projects & Integration Builds

Quick Access

Surgical Case Management & Clinical Procurement Integration
Capital Budget Integration (FBDI & ESS)
Invoice Order Attachment Integration
SFTP Migration — SCM OIC Refactor
SCM Middleware Error Notification Remediation
SCM Integration RCA & Stabilization (>99% Error Reduction)
ASN Receiving Integration
OCI PaaS Optimization
OpenWebUI LLM Deployment & Local LLM Integration

Oracle Integration Cloud (OIC) & Oracle Cloud Infrastructure (OCI)

Design and support integration solutions for Oracle Fusion Cloud ERP (SCM/FIN/HCM) via Oracle Integration Cloud (OIC) Gen3 under an change-control management governed delivery and change-control model. Each change documents business purpose, affected users, master data/configuration impact, environment-specific testing, approvals, and post-production follow-up. Technologies include: OIC Gen3, REST/SOAP APIs, Oracle ATP, BI Publisher, FBDI, ESS, PL/SQL, Java, XML, and SFTP.

Surgical Case Management & Clinical Procurement Integration

Engineered a bidirectional clinical procurement data exchange between an external EMR platform and Oracle Fusion ERP.
Outbound (ERP → EMR): Two BI Publisher interfaces — Item Master extract and PO lifecycle extract — published via SFTP to the EMR on automated schedules.
Inbound (EMR → ERP): OIC Gen3 orchestration ingesting purchase requisitions via SFTP, staged in ATP, enriched with ERP reference data via BI Publisher lookups, and submitted to Oracle ERP through FBDI and ESS job orchestration.
SOC 2-compliant fault handling, record-level ATP status tracking (NEW / SUCCESS / ERROR), and automated scheduling.
Coordinated with case schedule, case usage, and contract price data managed within the EMR platform to ensure end-to-end surgical case and procurement workflows.

Capital Budget Integration

OIC Gen3 scheduled orchestration ingesting daily capital project and budget extracts from an external EPM platform via SFTP.
Parses pipe-delimited data into ATP staging tables (project headers, task details, budget records).
Enriches records with Oracle reference data via BI Publisher reports: Project Manager EIN (AD Username lookup) and Department Code lookup — both required by Oracle Fusion Project Management FBDI.
Constructs Oracle-standard FBDI control files (PjfProjectsAllXface.zip and PjoBudgetsXface.zip) and submits to Oracle Fusion Project Control via ESS: Load Interface File for Import → Import Projects → Import Project Budgets.
De-duplication, version-aware budget update logic, partial-success fallout handling, batch-level email notifications, and SFTP-based FBDI archiving for audit.
Implements fault handling for SFTP, ATP, REST, and SOAP connections; monitors and confirms ESS job completion with wait/retry logic.

Invoice Order Attachment Integration

OIC Gen3 orchestration ingesting daily zipped invoice image containers (~500 invoices/day) from an external healthcare supply chain provider via SFTP.
Parses the index file to stage invoice metadata in ATP; resolves Oracle invoice IDs via BI Publisher (GetInvoice_Id report).
Dynamically constructs REST POST payloads with base64-encoded PDFs and invokes the Oracle Fusion Cloud Payables attachments endpoint.
Orphan PDF detection, record-level SUCCESS/ERROR status tracking, CSV error reporting, and SFTP archiving per batch run.
Global and local fault handling across SFTP, ATP, and REST connections.

SFTP Migration — SCM OIC Refactor

Refactored OIC file-polling orchestrations for SCM integrations to redirect inbound file acquisition from a decommissioned on-premises SFTP server to a vendor-hosted endpoint by inserting an OIC FTP/SFTP invoke step at the boundary layer.
Preserved all downstream ATP staging, FBDI construction, and ESS orchestration logic without rewriting existing integration flows.
Migrated seven outbound BI Publisher-scheduled SCM interfaces into OIC-managed schedulers to resolve a non-standard port incompatibility preventing direct BI Publisher SFTP delivery.
Maintained identical file contracts with the external system and added ad hoc BI Publisher output pickup logic for business continuity.

SCM Middleware Error Notification Remediation

During SFTP migration analysis, identified a silent-failure defect in OIC Shipping and Receipt import integrations where middleware errors produced no email notifications due to missing email body mappings in the shared error-handler switch condition.
Escalated during stand-up, received team approval to expand scope within the same delivery window, and implemented the targeted fix restoring full operational alerting for production middleware failures.
Resolved without redesigning the shared error framework, preserving all other integration flows downstream.

SCM Integration RCA & Stabilization (>99% Error Reduction)

Served as OIC SME for post-go-live stabilization following enterprise legacy ERP to Oracle Fusion Cloud ERP migration.
Gap Remediation: Performed comparative analysis of legacy SCM transaction semantics versus Oracle Fusion Receiving interface constraints, identifying undocumented dependencies in supplier site attribution, item-org UOM compatibility, and PO lifecycle synchronization.
Middleware Enhancements: Refactored inbound OIC integrations to introduce ATP-based canonical staging and BI Publisher-driven enrichment prior to FBDI/REST submission.
RCA Methodology: Established root cause analysis processes leveraging OIC instance logs, ESS request history, and ATP staging table queries to deterministically differentiate Oracle configuration defects from upstream data-quality violations.
Outcomes: Reduced monthly SCM receipt middleware errors from a peak of 5,672 (January 2023) to 7 (July 2025) — a >99% reduction. Reduced monthly Oracle cloud receiving errors from a peak of 24,383 (September 2022) to 385 (July 2025) — a ~98% reduction.
Operational Support: Authored production support playbooks enabling operations teams to locate, re-stage, and reprocess failed SCM transactions without manual ERP intervention.

OCI PaaS Optimization

Reduced infrastructure costs by automating OIC and ATP instance shutdown/startup schedules for off-hours.
Reduced the number of active OIC Connectivity Agents in test environments.
Maintained DR environments in standby mode, activating only for production backups or disaster recovery events.
Significantly lowered compute costs while preserving operational readiness across all environments.

OpenWebUI Deployment & Local LLM Integration

Maintaining and expanding an OpenWebUI platform for generative AI web integration, locally hosted small language models (SLMs), and cloud-based LLMs via a unified LangChain architecture. Key components include:

Tools: Groq Compound Systems Wolfram Alpha filter integration, Wolfram Alpha LLM API tool call, KJV Bible verse retrieval using Bible-kjv JSON library, Verse of the Day using BibleGateway Web-Service, and token optimization strategies through deterministic injection pipeline bypassing the LLM entirely for retrieval reducing context window consumption..
Models: Cloud instruct (GLM-4.5-Flash, Magistral-Small, Gemini-2.5-Flash, Groq-GPT-OSS-20B) and local GGUF (LiquidAI LFM2-8B, Qwen3:8b, Llama3.1:8b, DeepSeek-R1:8b, Mistral:7b) via Ollama and llama.cpp.
Stack: Python (FastAPI, REST), LangChain, WebSocket streaming, async/await, Docker, and Python venv/uv.
Features: RAG pipelines (Chroma + embeddings + reranking), MCP servers (mcpo / OpenAPI proxy), multi-agent workflows (A2A, ADK, ControlFlow), and Role-Pseudonymous Prompting (RPP) privacy protocol.
LLM API Integrations: Anthropic, OpenAI, Mistral.ai, Google GenAI, Groq, OpenRouter, Cohere, AWS Bedrock, HuggingFace, Ollama, and Perplexity.
Hardware: NVIDIA RTX 5070 Ti (12GB VRAM) for GPU-accelerated GGUF inference via llama.cpp (CUDA); AMD Ryzen 9 8940HX for CPU-only GGUF inference.
Benchmarks: Demonstrated that open-access LLMs with optimized RAG/Embedding pipelines achieve A+ parity with flagship models at 94.4% cost reduction.