Pablo Alessio

I built a custom non-thinking model, fed by real-time AI code generator agents, that automates complex travel itinerary decisions previously handled manually by travel specialists.

Before this new AI back end, every itinerary required human-crafted rules and field population; now, 85%+ of the itinerary is AI-driven, with agents interpreting natural-language business rules and itinerary context. I designed specialized background AI agents using the OpenAI platform (GPT-5 Mini) and Google Vertex AI (Gemini) to transform rule descriptions into executable Python logic for the rule engine, achieving a 90%+ time reduction in creating each travel itinerary and a 95%+ error reduction. The architecture is designed for multi-model AI orchestration. At the same time, the core back end runs on Python and FastAPI and integrates with a Node.js itinerary app for consumption and real-time itinerary outputs.

I designed and implemented an AI multi-agent orchestrator workflow from scratch, featuring a modular agent base and background workers. The orchestrator plans and routes tasks across base and specific agents with LangGraph and supervises related agent outputs. It integrates RAG to ground LLM prompts and unified tool adapters with cloud services with end-to-end observability. I also delivered a clean, scalable architecture with reusable agent functions, parallel tasking, and long-running jobs, minimal coupling via clusters, simplified connectors, client-ready FastAPI and React integration, a custom MCP server for real-time data ingestion into agents, and leveraging the Microsoft Agent Framework and Semantic Kernel.

I transformed a simulation learning platform into an AI-powered platform with a new cloud, DevOps, and app architecture. It had several long-standing issues that other engineers were not able to resolve. At the same time, I also doubled the speed performance of the apps, reduced Azure cloud computer costs by almost half, and stabilized the user experience with a more solid architecture, AI features, and monitoring tools. Implemented an advanced retrieval-augmented generation (RAG) process to leverage internal key business data with OpenAI models. Established prompt engineering and fine-tuning functionalities to adapt and retrain the models.

Developed an initial MVP of an ambitious AI agents platform that connects through ChatGPT agents and hosts with customized real-life actions using Semantic Kernel for advanced AI orchestration. I successfully implemented cutting-edge technologies to make the integration of AI language APIs with designed plug-in operations possible, enabling LLM-powered autonomous agents to execute complex, real-world tasks. Deployed a proof-of-concept marketplace for these agents, capitalizing on the enterprise automation trend with a monetized, subscription-based model. This platform showcases expertise in agentic workflows and the integration of generative AI into scalable, market-ready solutions.

Led an AI-focused modernization of the tech stack by systematically analyzing the current architecture and defining a high-impact software architecture and an AI roadmap. As the tech lead for AI strategy, I spearheaded the company's initiative to integrate next-generation artificial intelligence. My 1st step was a deep dive analysis of the existing technology stack to identify key opportunities and architectural challenges for AI adoption. I investigated a new AI strategy grounded in retrieval-augmented generation (RAG), enabling LLMs to contextually access internal data to drive smarter responses. I reviewed the running processes to orchestrate new agentic workflows, where autonomous agents leverage LLMs to carry out multi-step, goal-driven blueprints.
This strategic plan and detailed architecture now serve as the company's official roadmap for the software architecture, AI development, and innovation.

Architected a mission-critical system for a top American fintech company, establishing a new, scalable architecture. I defined the comprehensive cloud strategy and instituted a new framework with a robust set of best practices, architectural principles, patterns, and coding conventions. This approach provided the development team an unified development practices and set the technical roadmap for future growth, ensuring long-term stability and performance.

Designed and built a comprehensive fintech application for a major US startup, incorporating a full suite of financial tools. The app features sophisticated trading capabilities, pre-defined portfolio models, goal-based savings, and integrated full banking services. I engineered the app’s complex full-stack, which handles intricate financial calculations, third-party API integrations, and user-centric features. It delivers a highly engaging user experience with features like a rewards program, donation portal, eCommerce wishlist, and discounts originator.

Modernized a global cosmetics company’s legacy systems by designing and implementing cloud-native workflows on Azure. Built automated processes using Azure AI, Logic Apps, Functions, and Storage to streamline eCommerce order handling and internal operations.

Replaced outdated infrastructure with scalable, maintainable services, improving system reliability, reducing operational costs, and enhancing error handling and transparency across logistics workflows.

I created my personal website to bridge advanced technology with creativity and consciousness. It showcases my expertise in AI Engineering—focusing on LLMs and agentic workflows—alongside strategic insights into fintech and financial markets. It is a space dedicated to aligning cutting-edge innovation with mindful living and conscious investing.

I led the development of the Benchmark Submission Oversight (BSO) Tool, a framework I built from scratch to mitigate risk and increase transparency in Deutsche Bank's price submissions to external benchmarks. This independent supervisory application, integrated within the market risk management system, features automated alerts, escalations, and a streamlined workflow for Front Office and Market Risk Management collaboration. The BSO Tool successfully strengthened regulatory compliance and operational controls.

I engineered an end-to-end AML surveillance and case management tool for Santander Bank. The application features a real-time dashboard that provides analysts with a unified view of potential money laundering activities, streamlining the entire investigation workflow. This new system resulted in a significant improvement in the bank's risk mitigation capabilities, faster case resolution times, and a more proactive approach to identifying and reporting illicit financial activities.

I conceived and built Bindaloo, a social network platform designed to combat digital isolation by connecting people through shared real-world activities and hobbies. By focusing on shared activities, the platform enabled users to organize hundreds of their own meetups. This design prioritized genuine connection over digital consumption, resulting in a highly engaged user base that valued tangible offline relationships.

Co‑founded Kytasoft with a mission to compound client value through quality, automation, and measurable outcomes. I built a senior, product‑led team, lean engineering, and introduced cost‑aware designs and DevOps practices. This resulted in five successful product launches across different platforms, 100% on‑time delivery, 30–50% faster cycle times, 99.95% uptime on managed apps, and an ultimate user experience.

Unified the architecture and infrastructure of a manufacturer‑focused building modeling app, consolidating integrations behind a single API that connects CAD/BIM platforms and multiple front‑ends, cutting end‑to‑end model sync time by half. Added a versioned model pipeline with parametric components and automated validation, standardizing data exchange across platforms.

Re-architected a high‑throughput email marketing platform into a scalable, event‑driven system with advanced campaign management, real‑time analytics, and optimized marketing features. Achieved 45% faster send throughput, 25% lower infrastructure cost, and 99.9%+ uptime, delivering second‑level insights and precise campaign control.