Hi! I'm Wilfredo Sosa. I'm passionate about creating innovative solutions using the latest technologies in software development and artificial intelligence.
RadicalAI, New York, USA | August 2024 - Present
RadicalAI, New York, USA | June 2024 - Present
BoushTech, Delaware, USA | January 2024 - Present
BoushTech, Delaware, USA | July 2023 - December 2023
Developed a full-stack application designed to assist with three key services: Code Assistance (Chatbot), Data Structures (Documentation), NexaMind (AI Assistant for Software Architecture) and CodeXpert (AI Agent for improving code development and quality). The application enables users to generate code snippets and offers in-depth understanding of various data structures, including Arrays, Linked Lists, Stacks, Queues, Binary Trees, and Graphs. It features interactive learning modules powered by AI to enhance the educational experience. The frontend was built using NextJS, while the backend leveraged FastAPI, with AI capabilities integrated through multimodal LLMs, LangGraph with complex graphs, and the coexistence of GPT-4o-mini and Google Gemini for dynamic and intelligent content delivery.
Developed a cutting-edge platform that consolidates comprehensive and practical knowledge about Large Language Models (LLMs). The platform covers various dimensions, including Fundamentals, Prompt Engineering, Retrieval-Augmented Generation (RAG), Fine-Tuning, Model Deployment, Multimodal Models, and Security. It also offers real-world implementations using top-tier models like Google Generative AI, GPT-4o, and Claude Haiku. Users can explore interactive examples to understand how these principles are applied in practice, enhancing their grasp of modern AI techniques. The frontend was built with NextJS, while advanced AI capabilities are integrated through LangChain and state-of-the-art APIs.
Developed a full-stack application that provides a comprehensive suite of AI-powered writing tools with LangGraph. The application supports multi-language spelling checks, writing enhancement suggestions, addition of connectors for better flow, textual tone adjustments, and plagiarism detection to deliver superior writing assistance and improvement. Built using ReactJS for the frontend and FastAPI for the backend, the application leverages Google Generative AI and LLMs to offer intelligent, real-time feedback and guidance. The solution was deployed with a Docker Multi-Stage Build, ensuring robust performance, and was hosted on Vercel (frontend) and Render (backend).
Implemented a full-stack project for automating the retrieval of key concepts from different data types (PDFs, Word, PPTX, Youtube Videos and more). The frontend was built using NextJS with Tailwind and MUI to enhance UI/UX, while the backend leverages an AI pipeline that integrates FastAPI, LangChain, Vertex AI API, and YouTube transcript API for robust content extraction. In addition to the concept retriever, it features structured data study, semantic analysis, topic clustering, relation mapping, and sentiment analysis. Advanced RAG techniques like Self-Query Retrieval, Top-K retrieval, ChromaDB, Hybrid Search BM25, FAISS, EnsembleRetriever, HyDE, Multi-Query, RAG-Fusion, CoT prompting, and Parent-Document Retriever were implemented using GPT-4o-mini and Google Generative AI.
Built a full-stack application featuring four multimodal agents designed to support different phases of the Software Development Life Cycle (SDLC). These agents include the Refactoring Assistant, Documentation Generator Assistant, Multi-Agent Debugging Assistant, and LLM Application Development Assistant. The application integrates advanced AI pipelines powered by Google Generative AI and GPT-4o, allowing developers to streamline code refactoring, generate detailed documentation, debug issues using multi-agent collaboration, and build AI-powered applications. The frontend is developed with NextJS, and the backend leverages FastAPI for robust performance.
Created an innovative AI-powered platform aimed at enhancing the education process. LearnifyAI offers tools such as a Syllabus Generator, Rubric Generator, AI-Resistant Assignment Creator, and Personalized Task Generator. Leveraging Retrieval-Augmented Generation (RAG) and complex AI pipelines with models like GPT-4o and Google Generative AI, the platform tailors learning materials to individual students' needs, promoting more effective and resistant educational approaches. Built using NextJS for the frontend and integrated with advanced AI pipelines on the backend, LearnifyAI sets a new standard in AI-driven education solutions.
ReplyGen is an AI agent built to automate the generation of draft emails. By leveraging LangGraph, CrewAI, Gmail API, and Tavily, ReplyGen provides robust email generation and can pull relevant context when required knowledge falls outside the agent's scope. It uses the RAPTOR framework for advanced RAG (Retrieval-Augmented Generation), ensuring emails are accurate, context-aware, and ready for dispatch.
FlyOps comprises three AI multi-agents (Assistant, Analyst, Researcher) designed to automate the flight booking process. These agents query databases for flight availability, retrieve relevant information, and book flights based on payment conditions. The system is built with NextJS, React, Tailwind, FastAPI, LangChain, LangGraph, and CrewAI Streamlit UI, and incorporates Google Generative AI, SQLite, Uvicorn, Pydantic Schema, SQLAlchemy (ORM), and Tavily for a seamless, multimodal booking experience.
Implemented an AI-driven pipeline to automate the creation of various question types, including "Fill in the Blank", "Open Ended", "True/False", "Multiple Choice", "Relate Concepts", and "Math Exercises". The pipeline leverages the RAG (Retrieval-Augmented Generation) approach to extract context from diverse file formats such as PDFs, CSVs, Word documents, images, and YouTube videos. The frontend was built using NextJS and deployed on Vercel, while the backend integrated advanced technologies like LangChain, Vertex AI API, Streamlit UI, ChromaDB, and PyPDF to ensure robust question generation and efficient handling of different data sources.
Developed a full-stack application offering a range of AI-powered formatting and conversion tools, including Markdown Conversor, JSON Formatter, JSON to SQL, XML to SQL, Math Formulas to LaTeX, Translation, Info to Table, and Image Transcription. Built using NextJS for the frontend and FastAPI for the backend, it integrates LangChain for model orchestration and utilizes Google Generative AI and OpenAI for high-performance processing. The application was benchmarked for speed and accuracy, delivering optimized results. Deployed on Vercel with seamless performance and scalability for complex data transformations.
AI PPT Builder automates the creation of professional-grade slides by integrating a powerful tech stack that includes NextJS, React, Tailwind, FastAPI, LangChain, OpenAI GPT-4o-mini, and GoogleGenerativeAI. It utilizes Gradio and ChromaDB for Retrieval-Augmented Generation (RAG) capabilities, offering support for over 14 file types including PDFs, CSVs, PPTX, DOCX, and Google formats. This platform streamlines slide creation with advanced multimodal support for rich presentations.
Full-Stack app which supports the notes management with active and archived notes, as well as categories and efficient filters for optimal notes retrievement. It was made using NextJS, NestJS, PostgreSQL, Multi-Stage Build in Docker-Compose, compatibility with Linux/Windows and deployed in Vercel
Led a multidisciplinary team for extending the file support with more than 14 file types as well as enhancing the multimodal support for images and multilingual support. Additionally, we implemented a worksheet and syllabus generator; and optimized the key concepts retriever with efficient and state-of-art AI pipelines.
This portfolio contains more than 50 projects and lessons about Machine Learning fundamentals (Supervised, unsupervised, semi-supervised, and reinforcement learning)...
This repository contains several projects about Tensorflow fundamentals and best practices for NN, as well as advanced lessons about Neural Networks Architectures...
This repository contains more than 50 projects about LangChain with Google Generative AI, VertexAI, and GPT-4o-mini...
This course offers masterclasses about MIT CS and AI with real course materials from the university as well as projects with code that can support your learning process with important topics for Intro. to Algorithms, Advanced Data Structures, Advanced Graphs Management, Design and Analysis of Algorithms, Mathematics of Machine Learning, Supervised Learning, Unsupervised Learning and Intro. to Deep Learning.
This course offers masterclasses about the Master Degree in Artificial Intelligence, offered by the University of Texas at Austin, in outstanding areas like Machine Learning, Deep Learning and Planning, Search and Reasoning Under Uncertainty with more than 25 state-of-art projects about those topics.
This course contains several lessons about accomplishing a CS degree effectively with state-of-art subjects like Intro. to CS, Discrete Math, Programming Languages, OOP, Advanced Algorithms and Data Structures, Computation Theory, Analysis and Design of Algorithms, Artificial Intelligence, Databases, Networks, Competitive Programming and so forth.
This course contains important masterclasses for implementing DevOps in your current software lifecycle as well as MLOps and DLOps for integrating AI models to your apps and deploying them efficiently with the best practices.
This course contains several algorithms and models for Genetic Algorithms (GA), Evolution Strategies, Genetic Programming (GP), Differential Evolution (DE), Particle Swarm Optimization (PSO), Cultural Algorithms (CA), Estimation of Distribution Algorithms (EDAs), Memetic Algorithms (MA), Neuroevolution and Hybrid Evolutionary Algorithms.
This is an introductory course for learning Quantum Computing with integral theoretical lessons with Physics fundamentals as well as practical masterclasses for developing outstanding circuits, quantum gates and quantum algorithms with Cirq.
A systematic review focusing on the impact of LLMs in improving academic performance in higher education institutions.
This research examines how LLMs can be utilized to improve coding skills and enhance learning curves in programming education.
This study explores how Large Language Models (LLMs) can evaluate and analyze philosophical knowledge, enhancing the understanding of human consciousness through AI.
I'm always excited to connect with like-minded professionals and explore new opportunities. Feel free to reach out to me through the following platforms:
