Pipette tutorial by Viviana Márquez.
Watch the tutorial on X
July 2026 - Present
MTS, Developer Relations
at Liquid AI
Enabling developers to build fast, affordable, private AI that runs on-device, even offline, with Liquid AI's open-weight models. I combine technical depth, marketing expertise, and GTM vision to create practical resources, educate and support technical audiences, and connect developer engagement to business opportunities.
Community & Events
I lead community events and hackathons end to end, and coordinate C-suite events and conference participation. I develop technical demos and hands-on content for booths and hackathons, teach participants and the public how to use Liquid Foundation Models (LFMs), and provide technical support to help them build and troubleshoot.
My work also covers logistics, sponsorships, event promotion, and social media content capture. I send targeted outbound invitations to attract attendees who match our ideal customer profile (ICP), then hand off leads and opportunities to GTM for follow-up. I also represent Liquid AI as a speaker and panelist at developer and industry events.
Events I Organize
On-Device AI Hackathon: Cleared for Takeoff ✈️ Upcoming
Organizing the hackathon with Qualcomm, where I'll co-lead a hands-on workshop on running Liquid Foundation Models (LFMs) on IQ9 devices. I'm also managing logistics, event promotion, swag and social media design, targeted outbound invitations to attract attendees who match our ideal customer profile (ICP), and lead handoff to GTM.
Houston We Have No Cloud · #SFTechWeek
Organized the event with ZETIC. Co-led a hands-on workshop on running LFMs on-device and provided technical support to help participants build and troubleshoot. Managed the SF Tech Week submission, logistics, event marketing, swag and social media design, attendee follow-up, and lead handoff to GTM.
View workshop photos
Technical presentation at Houston We Have No Cloud Participants at the Houston We Have No Cloud workshop The Houston We Have No Cloud community group Liquid AI event banner at Houston We Have No Cloud Attendees networking at Houston We Have No Cloud The venue at Houston We Have No Cloud Outdoor demos at Houston We Have No Cloud Alura executive visit
Hosted 55 Brazilian senior executives visiting Silicon Valley. Delivered a technical presentation in Portuguese on Liquid Foundation Models and the benefits of running AI on-device. Coordinated the venue and team participation, followed up with attendees, and worked with GTM to arrange meetings.
Sponsorships & Conference Presence
NeurIPS 2026 Upcoming
SGLang Summit 2026 Upcoming
CoreWeave Fully Connected 2026
Tokens& · Long Horizon Agents Hackathon
AMD Advancing AI 2026
Speaking Engagements
DigitalOcean · Open Intelligence Summit
Panel: More Models, One Stack: How to Choose and Serve Open Models
Panelist alongside Peter Walker (OpenRouter), Alex Nails (RadixArk), and Victor Su Ortiz (MiniMax).
Technical Launches
I turn model and product releases into practical resources developers can use. Working across AI engineering and marketing, I build cookbook examples and technical demos, produce video tutorials, and create launch visuals and social content.
I manage Liquid AI's Discord community and its X and LinkedIn accounts.
| Launch | Description | Date |
|---|---|---|
| d1-ultra | An upgrade to d1 with stronger performance on knowledge-intensive tasks and more reliable confidence scores, built for demanding uses such as visual defect detection and damage assessment. | |
| LFM2.5-24B-A2B | An open-weight reasoning model with 24B total parameters and 2B active per token, built for fast, private on-device agents. | |
| d1-3B & d1-omni-600M | Open-weight decision models for edge deployment: d1-3B handles text and images, while d1-omni-600M adds audio support. | |
| d1 | A decision model that returns probabilities for classification, routing, and scoring tasks without generating text. | |
| LFM2.5-VL-DSpark | A lightweight draft model that accelerates LFM2.5 vision-language inference through speculative decoding while preserving output quality. | |
| Pipette | An open-source benchmarking platform for comparing foundation model speed, latency, and memory use on real edge devices. | |
| LongevityBench | An open benchmark for evaluating AI models on aging biology and longevity research. | |
| LFM2.5-Encoders | Compact text encoders for classification, language understanding, and retrieval, designed to run efficiently on local hardware. |