ERNEST GOMES

Professional Summary

AI Systems Architect and Prompt Engineer with hands-on experience designing, deploying, and managing multi-model AI orchestration platforms. Built production systems that coordinate 10+ large language models across automated research pipelines, code generation workflows, and real-time data analysis. Specializes in prompt engineering for complex multi-step reasoning, model evaluation and selection, cost optimization across AI providers, and autonomous agent design.

Technical Skills

Projects

AI Orchestration PlatformSolo architect and operator — Production system running 24/7 on cloud infrastructure

• Designed and deployed a multi-model AI orchestration platform that intelligently routes queries across 10+ LLMs based on task type, cost, and capability requirements
• Built a model audition framework that blind-tests 16 candidate models across standardized tasks, scored by independent AI judges, to assign the optimal model to each of 15+ specialized analyst roles
• Engineered a 5-phase advisory council system: question framing, parallel multi-model advisory generation, anonymized blind shuffle, independent peer review, and chairman synthesis — eliminating model bias through procedural design
• Created a multi-stage code generation pipeline where one model writes code, a second performs cold review, and a third conducts final audit — reducing defect rates through adversarial review
• Designed a 6-step research pipeline using model diversity (no model repeated within a pipeline) to produce fact-checked, multi-perspective research briefs
• Optimized API costs by 80%+ through strategic model selection, routing cheap models for simple tasks and reserving expensive models for high-stakes decisions
• Manages 117 automated scheduled jobs across data collection, health monitoring, and analytics

Real-Time Data Intelligence System290+ database tables | 130+ automated data sources | Multi-domain coverage

• Architected a comprehensive data pipeline ingesting real-time pricing, statistical, and sentiment data from 30+ external APIs and web sources
• Built automated data freshness monitoring with stale detection, error tracking, and alerting across all sources
• Designed a 13-desk analytical framework where each desk operates as an independent research unit with specialized AI analysts, dedicated data feeds, and structured output formats
• Implemented automated edge detection algorithms comparing model outputs against market prices across multiple domains
• Created a knowledge graph system linking entities, claims, and research findings with drift detection and version tracking

Tiered Memory Architecture for AI AgentsProduction memory system enabling long-term AI agent coherence

• Built an 11-phase memory system: auto-tagging, vector embeddings, hybrid search, entity versioning, claim tracking with drift detection, proactive recall with dynamic token budgeting, and episodic memory
• Implemented time-based memory tiers (Evergreen, Hot, Warm, Cool, Cold) with configurable decay and token budget allocation, validated through a 5-model panel review
• Designed retrieval-augmented generation (RAG) pipeline using FTS5 full-text search fused with cosine similarity on 384-dimension embeddings

Full-Stack Business Management PlatformComplete web application — Designed, built, and deployed in collaboration with AI

• Directed AI to build a 12-phase business management platform: inventory tracking, grading workflows, order management, financial reporting, customer management, and warehouse fulfillment
• Features include barcode scanning, multi-condition grading, purchase/sale tracking, profit analytics, pick-pack-ship warehouse workflow, and automated reporting
• Full relational database with 20+ tables, REST API backend, and responsive frontend deployed on cloud infrastructure

Multi-Platform Bot Network7 Telegram bots + 5 Discord bots — All running as managed background services

• Built and deployed 12 AI-powered chat bots bridging Telegram and Discord to various LLM backends
• Designed specialized bot roles: general assistant, analyst, tech support, and "fresh-eyes" judges that auto-wipe context between review sessions
• Implemented photo/screenshot OCR, voice transcription, session management, and graceful error handling

Automated Trading SystemsTwo independent algorithmic systems in production

• Designed prompt frameworks for an AI-assisted prediction market trading bot comparing model probabilities against live market prices to identify mispriced contracts
• Built monitoring and alerting for a forex trading system using technical analysis strategies
• Integrated real-time weather forecast APIs and sports data feeds for multi-domain signal generation

AI-Driven Security Audit Framework

• Designed a dual-audit security workflow: one AI model scans for vulnerabilities, a second independent model verifies findings and implements fixes
• Identified and remediated exposed services, overly permissive permissions, prompt injection vectors, and dependency vulnerabilities

Education

Electrical Engineering Certificate

Key Metrics