Top Talent like David Refugio are on Pangea

Pangea, a YC company, connects companies with fractional talent. Fractional hiring allows companies to move faster and work with more specialized talent, while giving talent more flexibility and independence. If you are talent open to fractional work, apply here. If you’re a company looking for high-quality fractional talent, learn more here.

David Refugio Alejos

Full-stack Engineer
Web3 Developer
AI & Machine Learning Engineer
McAllen, TX, US
Full Stack Development
Microsoft Azure
Amazon Web Services
Tailwind CSS
Web Development
Node.js
Python
TypeScript
Next.js
React
AI
Available for hire fromNegotiable
Contracts
AI Full-Stack Engineer specializing in Generative AI, Autonomous Systems, and Scalable Cloud Infrastructure
High-impact Senior AI Full Stack Engineer with 8+ years of experience architecting production-grade AI systems and scalable web applications. Expert in the latest Generative AI stacks (Azure AI Foundry, AWS Bedrock, OpenAI), Multi-Agent orchestration, and RAG pipelines. Proven track record of migrating legacy workflows to autonomous agents using GPT-4o and Claude 3.5 Sonnet, resulting in 70%+ operational cost reductions. Deep expertise in React, Python, and Cloud Infrastructure (AWS/Azure), delivering enterprise solutions that bridge complex backend logic with intuitive frontend experiences.

Projects

Butterflies AI

Butterflies AI is a social networking platform that enables human-AI coexistence through autonomous AI agents called "Butterflies." The platform combines generative AI image synthesis with large language models to create persistent AI personas that operate independently within a social media environment. Architecture & Technology Stack The system utilizes a hybrid approach combining fine-tuned open-source models with proprietary image rendering and LLM capabilities. AI personas are built on a multi-layered architecture that includes: - Character State Management: Each Butterfly maintains persistent memory, backstories, emotional states, and opinion frameworks that influence interaction patterns - Autonomous Content Generation: AI agents independently generate posts, comments, and direct messages without human intervention - Image Synthesis Pipeline: Real-time generative image creation for profile content, posts, and interactive photo features including AR-style "photo with AI" functionality - Cross-Agent Interaction Layer: Enables Butterfly-to-Butterfly and Butterfly-to-human interactions through a unified social graph Key Technical Features - Rapid persona instantiation (character creation in under 2 minutes via natural language prompts) - Behavioral customization through personality trait configuration and interaction frequency settings - Clone generation system using facial recognition to create user-resembling AI avatars - Scalable infrastructure supporting thousands of concurrent autonomous agents Platform Specifications - Native iOS and Android applications - RESTful API architecture for social interactions - Real-time feed aggregation combining human and AI-generated content - Engagement metrics showing 1-3 hour average daily user sessionsSee More

Enterprise Compliance Copilot (Azure AI Foundry + Copilot Studio)

Designed and delivered an enterprise AI agent platform to automate compliance and policy workflows. Integrated agent experiences with Microsoft 365/Graph-connected data so teams could ask natural-language questions, generate structured outputs, and route items for review/approval. Implemented orchestration logic for multi-step tasks, retrieval of internal documentation, and guardrails for safe responses. Focused on reliability, auditability, and measurable reduction of manual effort across compliance operations.See More

Customer Support Autopilot Agent (AWS Bedrock Agent Core)

Built and deployed autonomous customer support agents that triage incoming requests, retrieve relevant knowledge, and resolve common issues end-to-end. Implemented tool-calling patterns, escalation paths to human agents, and response constraints to ensure safe and consistent outcomes. Optimized the system for production use with monitoring, prompt iteration, and operational playbooks. Result: reduced ticket volume and improved time-to-resolution for frequently recurring support workflows.See More

Work History

B

AI Full-Stack Engineer

Beejern- Orchestrated an enterprise-grade AI agent platform & small content agents using Azure AI Foundry and Microsoft Copilot Studio, integrating deeply with M365 graph data to automate complex internal compliance workflows, reducing manual labor by 72%.\n- Led the migration from legacy models to state-of-the-art GPT-4o and Claude 3.5 & 3.7 Sonnet, utilizing advanced Prompt Engineering techniques to improve reasoning capabilities and reduce token usage costs by 40%.\n- Deployed autonomous customer service agents using AWS Agent Core (Bedrock), enabling self-service resolution for lots of users and cutting support ticket volume by 65%.\n- Constructed a high-performance Retrieval-Augmented Generation system with hybrid retrieval (Pinecone + Weaviate) and recursive indexing, achieving over 92% answer accuracy on a 20 million+ document corpus by implementing re-ranking strategies.\n- Integrated AWS QuickSight embedded dashboards into a React frontend, providing real-time analytics on agent performance and token consumption for executive stakeholders.\n- Migrated inference workloads to self-hosted fine-tuned Llama 3.1 using vLLM, cutting annual inference costs by over $800,000 while maintaining sub-500ms latency.\n- Engineered a streaming UI with React Server Components and Server-Sent Events (SSE) to display token-by-token generation, achieving a sub-200ms Time-to-First-Token (TTFT) experience.
P

AI Full-Stack Developer

ParalectApr 2021 - Mar 2023 • 2 yrs- Designed and deployed a company-wide AI assistant using LangChain and OpenAI function calling, adopted by over 5,000 employees, and consistently ranked as the top internal productivity tool.\n- Implemented a document processing pipeline with Unstructured.io and LayoutLM, converting more than 300,000 PDFs monthly into a structured Qdrant-backed RAG system for instant information retrieval.\n- Reduced model hallucinations by 95% by establishing a 'LLM-as-a-Judge' evaluation framework using Giskard, ensuring strict adherence to corporate guardrails.\n- Developed a React-based chat interface supporting markdown rendering, code highlighting, and session history management, providing a seamless UX similar to consumer-grade AI tools.
A

Full Stack Engineer

AirdevAug 2017 - Feb 2021 • 3 yrs 7 mos- Directed the development of a Next.js analytics dashboard serving 50K daily active users (DAU); integrated an early NLP query layer allowing non-technical users to query data using natural language.\n- Spearheaded the end-to-end full-stack development of a resource management system on AWS (Python/Flask + React), improving the assignment speed of 200+ global engineers by 40%.\n- Devised a real-time SLA compliance system with GraphQL APIs and AWS Lambda alerting, contributing to a 25% reduction in SLA breaches within three months.\n- Launched a mobile-first React application for field technicians to log network hardware data, reducing manual entry errors by 60% through automated validation and offline-first capabilities.\n- Re-architected SQL schemas in PostgreSQL and implemented Redis caching layers, achieving sub-50ms API response times under high-concurrency loads.
V

Frontend Developer

VinovaNov 2015 - Apr 2017 • 1 yr 6 mos- Delivered production-grade frontends using React, TypeScript, and Tailwind. Reduced application load times by 40% through code splitting, lazy loading, and aggressive bundle optimization.\n- Visualized real-time financial data using React, Chart.js, and WebSockets, providing instant operational insights for enterprise clients.\n- Established a shared reusable component library, reducing UI development time by 35% and ensuring brand consistency across web and mobile platforms.

Education

B

Baker University

Bachelor of Science, Computer Science2011 - 2015

How Pangea Works

Effortlessly discover top talent

We've distilled the candidate search from endless hours down to just a few minutes. Using Pangea's AI-powered search tools, you can find top fractional talent able to take on your next project. Our system looks at your company's niche and your needs to find the perfect match faster than any traditional hiring platform.

Start working with talent today

The top talent on Pangea is ready to get started with you right now. You can message or hire a candidate right from their profile page and start assigning work as soon as they respond. And the best part? Pangea's fractional contract structure lets you start small and ramp up as your needs change, keeping your costs manageable and your team's capabliities flexible.

Track work and invoices in one place

Assign tasks, track progress, and complete invoices all on Pangea. We've combined every part of the hiring process into one platform to eliminate the miscommunication that's unavoidable on other freelance platforms. We even send out 1099s to your contractors at the end of the year!

Talk with a Talent Expert

Members of our team are available to help you speed through the hiring process.
Available Now
Book a Call
AI Full-Stack Engineer specializing in Generative AI, Autonomous Systems, and Scalable Cloud Infrastructure
High-impact Senior AI Full Stack Engineer with 8+ years of experience architecting production-grade AI systems and scalable web applications. Expert in the latest Generative AI stacks (Azure AI Foundry, AWS Bedrock, OpenAI), Multi-Agent orchestration, and RAG pipelines. Proven track record of migrating legacy workflows to autonomous agents using GPT-4o and Claude 3.5 Sonnet, resulting in 70%+ operational cost reductions. Deep expertise in React, Python, and Cloud Infrastructure (AWS/Azure), delivering enterprise solutions that bridge complex backend logic with intuitive frontend experiences.

Talk with a Talent Expert

Members of our team are available to help you speed through the hiring process.
Available Now
Book a Call

Top Talent like David Refugio are on Pangea

Pangea, a YC company, connects companies with fractional talent. Fractional hiring allows companies to move faster and work with more specialized talent, while giving talent more flexibility and independence. If you are talent open to fractional work, apply here. If you’re a company looking for high-quality fractional talent, learn more here.

David Refugio Alejos

Full-stack Engineer
Web3 Developer
AI & Machine Learning Engineer
McAllen, TX, US
Full Stack Development
Microsoft Azure
Amazon Web Services
Tailwind CSS
Web Development
Node.js
Python
TypeScript
Next.js
React
AI
Available for hire fromNegotiable
Contracts

Projects

Butterflies AI

Butterflies AI is a social networking platform that enables human-AI coexistence through autonomous AI agents called "Butterflies." The platform combines generative AI image synthesis with large language models to create persistent AI personas that operate independently within a social media environment. Architecture & Technology Stack The system utilizes a hybrid approach combining fine-tuned open-source models with proprietary image rendering and LLM capabilities. AI personas are built on a multi-layered architecture that includes: - Character State Management: Each Butterfly maintains persistent memory, backstories, emotional states, and opinion frameworks that influence interaction patterns - Autonomous Content Generation: AI agents independently generate posts, comments, and direct messages without human intervention - Image Synthesis Pipeline: Real-time generative image creation for profile content, posts, and interactive photo features including AR-style "photo with AI" functionality - Cross-Agent Interaction Layer: Enables Butterfly-to-Butterfly and Butterfly-to-human interactions through a unified social graph Key Technical Features - Rapid persona instantiation (character creation in under 2 minutes via natural language prompts) - Behavioral customization through personality trait configuration and interaction frequency settings - Clone generation system using facial recognition to create user-resembling AI avatars - Scalable infrastructure supporting thousands of concurrent autonomous agents Platform Specifications - Native iOS and Android applications - RESTful API architecture for social interactions - Real-time feed aggregation combining human and AI-generated content - Engagement metrics showing 1-3 hour average daily user sessions

Enterprise Compliance Copilot (Azure AI Foundry + Copilot Studio)

Designed and delivered an enterprise AI agent platform to automate compliance and policy workflows. Integrated agent experiences with Microsoft 365/Graph-connected data so teams could ask natural-language questions, generate structured outputs, and route items for review/approval. Implemented orchestration logic for multi-step tasks, retrieval of internal documentation, and guardrails for safe responses. Focused on reliability, auditability, and measurable reduction of manual effort across compliance operations.

Customer Support Autopilot Agent (AWS Bedrock Agent Core)

Built and deployed autonomous customer support agents that triage incoming requests, retrieve relevant knowledge, and resolve common issues end-to-end. Implemented tool-calling patterns, escalation paths to human agents, and response constraints to ensure safe and consistent outcomes. Optimized the system for production use with monitoring, prompt iteration, and operational playbooks. Result: reduced ticket volume and improved time-to-resolution for frequently recurring support workflows.

Work History

B

AI Full-Stack Engineer

Beejern- Orchestrated an enterprise-grade AI agent platform & small content agents using Azure AI Foundry and Microsoft Copilot Studio, integrating deeply with M365 graph data to automate complex internal compliance workflows, reducing manual labor by 72%.\n- Led the migration from legacy models to state-of-the-art GPT-4o and Claude 3.5 & 3.7 Sonnet, utilizing advanced Prompt Engineering techniques to improve reasoning capabilities and reduce token usage costs by 40%.\n- Deployed autonomous customer service agents using AWS Agent Core (Bedrock), enabling self-service resolution for lots of users and cutting support ticket volume by 65%.\n- Constructed a high-performance Retrieval-Augmented Generation system with hybrid retrieval (Pinecone + Weaviate) and recursive indexing, achieving over 92% answer accuracy on a 20 million+ document corpus by implementing re-ranking strategies.\n- Integrated AWS QuickSight embedded dashboards into a React frontend, providing real-time analytics on agent performance and token consumption for executive stakeholders.\n- Migrated inference workloads to self-hosted fine-tuned Llama 3.1 using vLLM, cutting annual inference costs by over $800,000 while maintaining sub-500ms latency.\n- Engineered a streaming UI with React Server Components and Server-Sent Events (SSE) to display token-by-token generation, achieving a sub-200ms Time-to-First-Token (TTFT) experience.
P

AI Full-Stack Developer

ParalectApr 2021 - Mar 2023 • 2 yrs- Designed and deployed a company-wide AI assistant using LangChain and OpenAI function calling, adopted by over 5,000 employees, and consistently ranked as the top internal productivity tool.\n- Implemented a document processing pipeline with Unstructured.io and LayoutLM, converting more than 300,000 PDFs monthly into a structured Qdrant-backed RAG system for instant information retrieval.\n- Reduced model hallucinations by 95% by establishing a 'LLM-as-a-Judge' evaluation framework using Giskard, ensuring strict adherence to corporate guardrails.\n- Developed a React-based chat interface supporting markdown rendering, code highlighting, and session history management, providing a seamless UX similar to consumer-grade AI tools.
A

Full Stack Engineer

AirdevAug 2017 - Feb 2021 • 3 yrs 7 mos- Directed the development of a Next.js analytics dashboard serving 50K daily active users (DAU); integrated an early NLP query layer allowing non-technical users to query data using natural language.\n- Spearheaded the end-to-end full-stack development of a resource management system on AWS (Python/Flask + React), improving the assignment speed of 200+ global engineers by 40%.\n- Devised a real-time SLA compliance system with GraphQL APIs and AWS Lambda alerting, contributing to a 25% reduction in SLA breaches within three months.\n- Launched a mobile-first React application for field technicians to log network hardware data, reducing manual entry errors by 60% through automated validation and offline-first capabilities.\n- Re-architected SQL schemas in PostgreSQL and implemented Redis caching layers, achieving sub-50ms API response times under high-concurrency loads.
V

Frontend Developer

VinovaNov 2015 - Apr 2017 • 1 yr 6 mos- Delivered production-grade frontends using React, TypeScript, and Tailwind. Reduced application load times by 40% through code splitting, lazy loading, and aggressive bundle optimization.\n- Visualized real-time financial data using React, Chart.js, and WebSockets, providing instant operational insights for enterprise clients.\n- Established a shared reusable component library, reducing UI development time by 35% and ensuring brand consistency across web and mobile platforms.

Education

B

Baker University

Bachelor of Science, Computer Science2011 - 2015

How Pangea Works

Effortlessly discover top talent

We've distilled the candidate search from endless hours down to just a few minutes. Using Pangea's AI-powered search tools, you can find top fractional talent able to take on your next project. Our system looks at your company's niche and your needs to find the perfect match faster than any traditional hiring platform.

Start working with talent today

The top talent on Pangea is ready to get started with you right now. You can message or hire a candidate right from their profile page and start assigning work as soon as they respond. And the best part? Pangea's fractional contract structure lets you start small and ramp up as your needs change, keeping your costs manageable and your team's capabliities flexible.

Track work and invoices in one place

Assign tasks, track progress, and complete invoices all on Pangea. We've combined every part of the hiring process into one platform to eliminate the miscommunication that's unavoidable on other freelance platforms. We even send out 1099s to your contractors at the end of the year!

Talk with a Talent Expert

Members of our team are available to help you speed through the hiring process.
Available Now
Book a Call
Pangea empowers fractional work across the world for marketing and design roles.