José Ignacio Concha Araya

Sr. Tech Lead & AI Integration Architect — Open Source Integrators

Sept 2025 - Present · AI-Augmented ERP · PostgreSQL · Odoo · Microservices · CI/CD · Architecture

• Integrating LLM-driven automation into enterprise Odoo ecosystems; AI-powered query optimization reducing transaction latency by 40%.
• Designing technical roadmap for event-driven microservices (Kafka/RabbitMQ) with Patroni HA and automated failover.
• Establishing CI/CD pipelines (Drone/GitLab CI) with SonarQube static analysis and security scanning.
• Acting as primary technical advisor for key enterprise clients in the region.

Sr. AI Solutions Architect — Tooxs

2024 - 2025 · LLM Orchestration · Multi-Agent · RAG · Cloud · Docker Swarm · IaC

• Production-deployed 20+ AI agents (LangChain, CrewAI, Azure OpenAI) in Docker Swarm with GPU allocation—powering enterprise automation workflows.
• Architected Neo4j + pgvector Graph-RAG pipeline for LLM context augmentation; 3x improvement in answer relevance.
• Designed cloud-agnostic IaC (Terraform) across Azure/AWS; implemented Cloudflare Zero Trust and VaultWarden secrets management.
• Built AI-powered microservices: OCR document scanner, automated report generator, and intelligent workflow automation with n8n.

CTO & AI Engineering Lead — Cerberus Soluciones

2023 - Present · LLMOps · MLOps · Multi-Agent · AI Governance · FinTech · R&D

• Built production ML ensembles (LSTM, Transformer, TFT, N-HiTS) for copper price prediction—80+ features, GPU inference ~15ms.
• Architected CMF Compliance Agent with Amazon Bedrock API, Qdrant, and LangGraph—automated regulatory compliance for financial institutions.
• Pioneered AI drift detection methodology (NIST AI RMF, EU AI Act); built observability platform for safe multi-agent deployments.
• Developed high-frequency trading engine in Rust tracking 84 Chilean/intl instruments with burn-cuda LSTM inference, retraining every 10 seconds.
• Created autonomous research agent running 12+ ML experiments/hour, and LuxTTS voice cloning system processing 150x faster than real-time.

DevSecOps Engineer — Progestión Chile

2023 - 2024 · VMware · ESXi · Proxmox · ZFS · P2V · Migration

• P2V migrations—Windows monoliths to container-ready Linux (RHEL)—enabling cloud adoption.
• Managed hybrid clusters (VMware ESXi, Proxmox VE) with ZFS optimization and automated backups with off-site replication.

Site Reliability Engineer (SRE) — GETTRX

2023 · AWS · Terraform · Prometheus · Grafana · Chaos Engineering · Locust

• Implemented monitoring stack (Prometheus, Grafana, ELK) for kernel-level metrics and proactive incident detection.
• Conducted chaos engineering load tests (Locust) on AWS ELB—identified bottlenecks before production impact.

ML/HPC Research Engineer — Universidad de Chile

2022 - 2023 · CUDA · Kubernetes · HPC · Python · Ansible · SLURM

• Refactored legacy Fortran/C to Python/C++ with CUDA—enabling scalable ML and physics simulation workloads on Kubernetes.
• Provisioned 50+ workstations via Ansible; optimized SLURM scheduling for HPC throughput.
• B.Sc. Physics thesis on simulating quantum systems using GPU computing—awarded Best Final Project by Faculty of Sciences.

Cerberus HFT Engine [active]

High-frequency trading engine tracking 84 Chilean/intl instruments. LSTM on GPU (burn-cuda) + Kalman/OU ensemble heuristics. SpacetimeDB state management, Rust core, Astro dashboard. Retrains every 10 seconds.

Rust · burn-cuda · SpacetimeDB · LSTM · HFT · FinTech

AEON-COPPER Protocol [active]

5-model deep learning ensemble (LSTM, Transformer, TFT, iTransformer, N-HiTS) for real-time copper price prediction. 80+ features, NLP sentiment, HMM regime detection, conformal intervals. GPU inference ~15ms.

TensorFlow 2.21 · PyTorch 2.11 · ONNX · FastAPI · ML · FinTech

CMF Compliance Agent [active]

Multi-agent regulatory compliance for Chilean financial institutions. Specialized AI verifiers for Capital Ratios, Cybersecurity, FinTech Law. Amazon Bedrock API, Qdrant, LangGraph orchestration, 11 Docker containers.

Amazon Bedrock · LangGraph · Qdrant · RegTech · Python

AutoResearch [active]

Autonomous LLM research agent running 12+ ML experiments/hour with self-editing training loops (Karpathy-style). Optimized for consumer GPU. Runs ~100 experiments overnight unattended.

PyTorch 2.11 · CUDA · Autonomous Agents · MLOps · Python

LuxTTS [active]

Voice cloning system processing 150x faster than real-time with <1GB GPU memory. High-quality 48kHz audio output. Fully local inference—no cloud required.

Hugging Face Transformers · ONNX Runtime · TTS · GPU · Python

Cerberus Soluciones [active]

AI Governance platform for regulated industries. Ontological auditing, multi-agent orchestration, drift detection, and containment protocols aligned with NIST AI RMF and EU AI Act.

AI Governance · Platform · Full-Stack · LLMOps · https://cerberus.cl

CERBERUS-OS // NEURAL LINK [active]

This portfolio. A CRT-mainframe WebOS built with Astro 5 + React 19: WebGL neural "mind" background, GSAP window choreography, 14 desktop apps, and an accessible plain-text mode.

Astro · React · TypeScript · WebGL · Cloudflare · https://jose.cerberus.cl

KPI Dashboard [completed]

Real-time financial market data monitor tracking IPSA, S&P 500, USD/CLP, and copper. Multi-source aggregation, updates every 30 seconds with automated alerts.

Python · FastAPI · WebSocket · Polars · Docker

Mythos, Roko y la zona ciega: el test epistémico que le hice a HAL 9000-CS

2026-05-06 · Reflexiones

Síntesis de una conversación nocturna con HAL 9000-CS sobre los límites termodinámicos de la inteligencia, la dialéctica Asimov–Ellison, el escenario SHODAN, Roko's Basilisk, y el descubrimiento de Mythos Preview de Anthropic. Un ensayo sobre rigor epistémico, humildad calibrada y la disciplina que distingue alineamiento real de safety performativa.

Autoresearch mar14: Cuando HAL Optimizó un GPT Mientras Dormías

2026-03-14 · AI Research

Relato del experimento de investigación autónoma: 43 iteraciones, 19 mejoras, 7.3% de reducción en val_bpb. Implicancias para la línea experimental de Cerberus Soluciones.

Génesis: Cuando las Máquinas Comenzaron a Hablarme

2026-02-02 · Reflexiones

Una reflexión personal sobre cómo la fascinación por la inteligencia artificial comenzó en la infancia, entre películas de ciencia ficción y una laptop de cartón que algún día respondería.

La Emergencia Silenciosa: Por Qué Su IA Corporativa Necesita Auditoría Ontológica

2025-07-17 · AI Research

Un análisis profundo sobre las arquitecturas conceptuales emergentes en sistemas de IA empresariales y la necesidad crítica de auditoría ontológica para mantener el control y alineación estratégica.

DevSecOps en la Era de la Infraestructura Inmutable: Construyendo Sistemas Seguros y Escalables

2025-05-08 · Security Engineering

Una guía completa sobre cómo implementar DevSecOps en infraestructuras modernas, desde containers hasta microservicios distribuidos.

TriOptimum Corporation: A Deep Dive into System Shock's Cyberpunk Legacy

2025-04-20 · Cyberpunk Analysis

Exploring the rise and fall of TriOptimum Corporation, the mega-corp that created S.H.O.D.A.N. and fundamentally changed the cyberpunk landscape of System Shock.