Bad Orange (DaddySounds)
2026–nowBackend for a voice companion: text in, voice out.
Experience
Fourteen years, three eras.
Backend systems, then machine learning, then AI that talks back. Here’s all of it, newest first.
2023–now
AI engineering
Large language models arrived and I went independent. Agents, retrieval, voice and evals, for clients from legal tech to restaurants.
12 rolesIndependent since 2023. Most of this work ran through DataSmarts, my studio.
Liveday
2026Multilingual immigration Q&A and governed retrieval.
Palm Island
2026Charming Chad: live voice practice with an AI referee.
Compliance 12.5% → 82.3%Case study →
Julian Mulvey · SpotFeed
2025–2026Turned 40k FCC filings into a searchable ad-spend platform.
~$77M trackedCase study →
Crescendo Creative
2025–2026Forge: eight AI agents the team works with in Slack.
DesktopClassroom
2025–2026Orchestration for a multimodal tutor with LangGraph and DSPy.
United Claims Specialists
2025Automated lead reporting and built a retrieval assistant.
5 h/week → under 2 min
JIA Nomads
2024–2026Email intelligence pipeline and agents across four chat apps.
3 min → under 20 s
Padrino’s
2024–2025Bilingual staff assistant across five restaurants.
~75% fewer manager interruptions
Cortado
2024–2025Rebuilt the retrieval behind AI guest messaging, plus an MCP server.
Replies: minutes → under oneCase study →
Komorebi
2024An agent that answers plain-English questions with SQL.
Planning time −20%
ContractPod
2023–2024Fine-tuned Llama 3 and Mistral inside isolated environments.
Matched GPT-4
2019–2023
ML engineering
Computer vision, NLP and forecasting in production, for pharma, legal and retail. I wrote a book along the way.
3 rolesMahisoft
2020–2023Rebuilt a named-entity recognition system; computer vision prototypes.
Entity recognition 4× faster
Blue Orange Digital
2019–2022Document tagging, demand forecasting and an OCR rescue.
Beat human reviewers by 10%
Freelance
2019People counters and a depth-camera age gate.
10+ locations
2014–2019
Software engineering
Backend systems and microservices for insurance, fintech and adtech. Where I learned what production means.
4 rolesMahisoft
2017–2020Authentication and notification services in Go and Java.
A SQL job: 2 h → 9 min
Wivo
2016–2017Scala and Clojure microservices, Docker sandboxes for every developer.
Setup: days → minutes
Mesfix
2015–2016Moved to microservices and set up the team’s first CI/CD pipeline.
S4N
2014–2015Secure storage and client onboarding for an insurer.
Onboarding: 1 week → ~3 hours
Credentials
On paper.
Education
BSc, Computer Engineering
Universidad Simón Bolívar · 2008–2013
Book
TensorFlow 2.0 Computer Vision Cookbook
Packt · 2021 · sole author
Certificates
Seven in machine learning
Stanford (Coursera), deeplearning.ai, Udacity, UC San Diego
Languages
English and Spanish
Working in both, every week