I build AI systems that connect models to real-world decisions — and explore how they can be turned into useful tools.
🧠 Portfolio
https://dosorio79.github.io/portfolio
Selected projects, write-ups, and demos live there — this GitHub hosts the implementation.
- 📊 Decision-focused data science (pricing, uplift modeling, marketing systems)
- 🤖 GenAI applications (RAG pipelines, agentic workflows, local-first setups)
- 🧠 NLP for customer feedback and operational insights
- 🚀 End-to-end AI systems (APIs, deployment, evaluation, monitoring)
-
🤖 uLearn — Agentic micro-learning platform
Explores how AI systems can become structured, usable learning tools with validation and telemetry -
🧩 RAG pipelines with local LLMs
Practical exploration of retrieval systems using FastAPI, ChromaDB, and Ollama -
🧠 AI coding & agent systems
MCP tools, coding agents, and collaborative coding workflows -
👁️ CompVis — industrial safety face recognition
Repo: https://github.com/kolapally/computer_vision
Demo: https://compvis.streamlit.app/
More context and walkthroughs in the portfolio.
- Agentic AI systems (practical > hype)
- LLMOps and local model deployment
- Evaluation of AI systems (RAG, decision impact)
- Data products with measurable business value
GenAI systems, RAG pipelines, experimentation, pricing analytics — or actomyosin cytoskeletons and CRISPR.
- LinkedIn: https://linkedin.com/in/dosorio
I started in academic cell biology, studying the actomyosin cytoskeleton.
That experience shaped how I approach systems today: hypothesis-driven, experimental, and grounded in real mechanisms.
I now apply the same mindset to data science and AI — building systems that are not only accurate, but useful, testable, and deployable.
Python · FastAPI · Docker · Vector DBs · RAG · Local LLMs · Polars · scikit-learn
⚡ Fun fact: My first PC had 40 MB storage and 1 MB RAM — constraints were the original teacher.