{"about_me":{"education":{"header":"EDUCATION","values":[{"coursework":["Coursework Programming","Datastructure and Robotics"],"degree":"BS Computer Engineering","school":"Collegio De Montalban","year":"2014"}]},"outside_work":{"header":"ME OUTSIDE WORK","values":["Dad","Bassist","Musician","Woodworker","Luthier","Cobbler","Cat Lover"]}},"contact_me":{"email":"jayrcastro.py@gmail.com","github":{"display":"jayrcastropy","url":"https://github.com/jrcastropy"},"header":"CONTACT DETAILS","linkedin":{"display":"www.linkedin.com/in/jayrcastro/","url":"https://linkedin.com/in/jayrcastro"}},"experiences":{"header":"Work Experiences","values":[{"company":"Moneydolly - AI Agent (Finn)","description":"Built Finn, a multi-autonomous AI Agent for Moneydolly's fundraising platform.<br>Finn leverages MCP server for tools, enabling users to create fundraising campaigns instantly, send messages, escalate issues, and answer questions using RAG.<br>Integrated Slack notifications for campaign managers and implemented phone call capabilities for seamless user communication.","link":"https://moneydolly.com/","title":"AI Engineer","years":"2025-2026"},{"company":"Cooks Florist - AI Phone Assistant (Kayla)","description":"Designed Kayla, a SOL-style AI voice agent that answers and places calls, sends texts/emails, and guides customers through Cook's Florist orders.<br>I architected an API + MCP toolchain for knowledge, product, deliverability, and payment handoffs so every sub-300-character reply stays accurate, warm, and compliant.","link":"https://cooksflorist.com/","title":"AI Engineer","years":"2026"},{"company":"Team Floral","description":"AI Assistant Development & Website Creation for TeamFloral<br>I designed and developed Rosie, an advanced AI assistant for TeamFloral, that automates report generation, internet searching, and web scraping. Rosie processes data using MCP (Machine-Comprehension Processing) and RAG (Retrieval-Augmented Generation), ensuring efficient and accurate information retrieval. The assistant is integrated with MongoDB for robust data storage and management.<br>In addition to the backend and AI work, I was responsible for the creation of both the front-end and back-end of the website, ensuring a seamless and responsive user experience.<br>As part of the AI development, I fine-tuned a machine learning model on a proprietary dataset called 'Flower and Profits', which is used for advanced analysis and reporting. Furthermore, I implemented and trained a Prophet model, enabling the prediction of sales trends for upcoming months, empowering the business to make data-driven decisions.","link":"https://teamfloral.com/","title":"AI Engineer","years":"2024 to 2025"},{"company":"New Age Media Group","description":"Developed an advanced audio matching system to identify commercial audio within radio station streams, ensuring accurate playback timing. Implemented a comprehensive Text-to-Speech (TTS) function utilizing AWS Lambda for seamless audio integration.Developed an advanced audio matching system to identify commercial audio within radio station streams, ensuring accurate playback timing. Implemented a comprehensive Text-to-Speech (TTS) function utilizing AWS Lambda for seamless audio integration.","link":"https://www.linkedin.com/company/namg/","title":"AI Engineer","years":"2023 to 2024"},{"company":"Zero Company Marketing Performance, USA","description":"Built an AI Chief Marketing Officer (AICMO) that provides different kinds of marketing tools, from building a marketing plan to creating landing page to ad copies with one push of a button.","link":"https://www.zerocompany.com/","title":"AI Product Engineer","years":"2021 to 2023"},{"company":"Boredhumans.com, USA","description":"Created bunch of AI web applications for boredhumans.com. The website was originally created for people who want to try AI Applications without coding or hassle. Trained models including CV, Text2Audio, Audio2Text, NLP, and LLMs.","link":"https://boredhumans.com/","title":"Senior Machine Learning Engineer","years":"2021 to 2023"},{"company":"MetrocityAI Start Up, Philippines","description":"Built Hiring AI that reads CVs using OCR and NLP to compare with Job Descriptions. Also added an Audio-to-Text feature to analyze applicant responses in interviews.","link":"https://metrocity.ai/","title":"Chief Technology Officer / Machine Learning Engineer","years":"2019 to 2022"},{"company":"Thobias.com","description":"Developed Photoshoot Selection using Pose Estimation and Learning to Rank. Built a Raw Video to Production pipeline using GANs and Pose Estimation. Also developed Deep Fashion for clothing/accessory detection.","link":"https://thobias.com/","title":"Machine Learning Developer","years":"2021 to 2022"},{"company":"Job Tensor UK / DE","description":"Fine-tuned transformer models to generate occupations from job posts. Used NLP to extract company locations from job descriptions.","link":"https://jobtensor.com/","title":"Data Scientist / Machine Learning Engineer","years":"2020 to 2021"},{"company":"Monster Group UK LTD., UK","description":"Created a web app to detect invoice discrepancies using AI.","link":"https://www.youtube.com/watch?v=9zaNlxsNuL8","title":"Software Developer / Machine Learning Engineer","years":"2020 to 2021"},{"company":"BlickLabor Start Up, Germany","description":"Built an ML software for automated checkout detecting grocery products in supermarkets.","link":"https://www.youtube.com/watch?v=YE1oZ6F6sug","title":"Software Developer / Machine Learning Engineer","years":"2019 to 2020"},{"company":"Reed Elsevier Share Services, PH","description":"Built process automation, including 'Nimmer QC Automation' which reduced processing from 1 month to 3 days.","link":"https://www.reedelsevier.com.ph/","title":"Content Automation Specialist / RPA Developer","years":"2017 to 2020"}]},"projects":{"header":"Other Projects","values":[{"description":"AI chatbot platform offering personalized, judgment-free conversations with AI companions. Offers both free and subscription tiers.","link":"https://rolepley.ai/home","role":"LLM / NLP / ML Engineer","title":"Rolepley.ai"},{"description":"Web app that improves users\u2019 communication, vocabulary, and argumentation skills.","link":"https://articulacy.app/","role":"Machine Learning Engineer","title":"Articulacy.app"},{"description":"Job Ads website using GPT-3 to neutralize gender bias. Computes applicant-job similarity using Transformer embeddings and vector databases.","link":"https://equal.ai","role":"Machine Learning Engineer","title":"Equal.AI"},{"description":"Chatbot for PoliceAuctions.com using GPT-3 prompt engineering with memory and embeddings.","link":"https://sitebots.ai/","role":"Machine Learning Engineer","title":"PoliceAuction Chatbot"},{"description":"Project for Yannic Kilcher using OpenAI Clip and BigGAN to auto-generate music videos based on lyrics.","link":"https://www.youtube.com/watch?v=rR5_emVeyBk","role":"Machine Learning Developer","title":"OpenAI Clip and BigGAN Music Video"},{"description":"One-click app that scrapes Reddit posts, turns them into creative videos, and auto-uploads to YouTube.","link":"https://www.youtube.com/watch?v=kn_kasrDSDk","role":"Machine Learning Developer","title":"Reddit Automation"},{"description":"Trained an ML model on MNIST from scratch using only Numpy without any ML frameworks.","link":"https://github.com/jrcastropy/ML_using_only_Numpy_MNIST_Dataset/blob/master/MNIST%20Dataset%20using%20numpy.ipynb","role":"Machine Learning Developer","title":"Machine Learning with only NUMPY (MNIST DATASET)"}]},"skills":{"header":"Skills","values":[{"header":"Python","values":["MediaPipe CV","OpenCV","Tensorflow","Pytorch","MXNet","NLTK","Spacy","Gensim","HuggingFace","Numpy","Pandas","Flask","Boto3 (AWS)","Matplotlib","Requests","BeautifulSoup","RestAPI","Web Scraping","PyQT","Tkinter","Pymongo","Scrapy","LangChain","LangGraph","LlamaIndex","Pydantic","Scikit-learn","Keras","FastAPI","Streamlit","Gradio"]},{"header":"Machine Learning","values":["Deepseek","Gemma","OpenAI (API)","OpenAI Function Calling","OpenAI Tools Calling","Llama","Llama2","GPT-J","GPT2","GPT3","GPT4","LORA","VICUNA","FLAN-T5","Text Embeddings/Vectors","Sentence Transformers","Semantic/Document Search","Stablediffusion","Facial Recognition","Pose Estimation","Image Recognition / Classification","Object Detection","YoloV3/V4","EfficientNet","DeepLabV3","Facebook Detectron","Transformers HuggingFace","Attention Models","Language Translation Models HuggingFace","Neural Networks","Computer Vision OpenCV","GAN (General Adverial Network)","BigGan","StyleGAN","Reinforcement Learning","OpenAI RL Environment GYM","OpenAI Dall-E","OpenAI Clip","Deep Tagging","Deep Fashion","Claude / Anthropic API","Claude Code","Claude AI Agents","MCP Server","OpenClaw","Text to Speech (TTS)","Speech to Text (STT)","Whisper","Twilio","VAPI","Gemini","Mistral","CrewAI","AutoGen","RAG (Retrieval Augmented Generation)","Vector Databases (Pinecone, ChromaDB, Weaviate)","MLOps","MLflow","Prompt Engineering","Fine-tuning LLMs","ONNX","Agentic AI","JAX","Google Vertex AI","Amazon SageMaker","Azure Machine Learning","IBM Watson Studio","Comet ML","Weights & Biases (W&B)","DVC (Data Version Control)","Kubeflow","Apache Airflow","Dagster","Ray","Dask","OpenAI Swarm","Microsoft Semantic Kernel","Haystack","Rasa","Llama 3","Claude 3.5 Sonnet","Cohere API","Qwen","Phi-3","SDXL (Stable Diffusion XL)","ControlNet","Flux AI","ComfyUI","Diffusers (HuggingFace)","TensorRT","OpenVINO","BentoML","Triton Inference Server","vLLM","Text Generation Inference (TGI)","Ollama","OpenRouter","Groq API","Together AI","Replicate","RunPod","RLHF (Reinforcement Learning from Human Feedback)","DPO (Direct Preference Optimization)","QLoRA","PEFT (Parameter-Efficient Fine-Tuning)","DeepSpeed","llama.cpp","Multimodal AI","Vision Language Models (VLMs)","LLaVA","GPT-4 Vision","LangSmith"]},{"header":"Cloud","values":["AWS (AMAZON WEB SERVICES)","AWS Lambda","AWS Transcribe","AWS Rekognition","AWS Sagemaker","AWS Polly","AWS Bedrock","AWS Route 53","AWS S3","AWS Elastic Beanstalk","AWS EC2","AWS Codepipeline","AWS Lightsail","AWS API Gateway","GCP (GOOGLE CLOUD PLATFORM)","GCP Cloud Functions","GCP Cloud Storage","GCP Charts","GCP APIs","Supabase","Supabase Edge Function"]},{"header":"Programming Languages","values":["ReactJS","Typscript","Deno","NodeJS","Jquery","Javascript","Javascript ES6","HTML","CSS","Bootstrap","MySql","Docker","Git","Ajax","GraphQL","Shopify Liquid","MongoDB","PostgresSQL","VB.NET","Java","C++"]}]}}