Private intelligence

Private, offline AI you own.

FlorenciaAI is a self-hosted assistant that runs on hardware you own — fully offline, on open-weight models, with no dependence on a cloud frontier model. We are now accepting pilot inquiries for private, on-premise AI servers for law firms, medical practices, and accountants. The long-term goal is a private operating system for a fleet of off-the-shelf humanoid robots, centrally controlled from that same kind of server.

Deployment
On-premises, private
Models
Open-weight, local
Status
Prototype · pilots open
Program
Member of NVIDIA Inception
01

The problem is dependence.

Most assistants today live in someone else’s cloud. The conversation, the memory, the tools, and — before long — the commands that move a robot all travel to a frontier model you do not operate and cannot keep.

That arrangement is convenient. It is a poor foundation for a system meant to stay in the house, remember what matters, and eventually coordinate machines on a private network.

That gap is sharper still for lawyers, doctors, and accountants, who often cannot put client or patient data in a cloud they do not control. FlorenciaAI starts from a different premise: the intelligence should belong to the person who uses it, run on hardware they own, and keep working when the outside world is not invited in.

Cloud dependence

Capability is rented from a provider. If the service changes, goes dark, or raises the price, the assistant goes with it.

Privacy leaves the room

Mail, research, household context, and future robot control are too intimate to hand to a model running elsewhere by default.

No private network

A house, a workshop, and a set of robots need a local mind — not a session in a browser tab on someone else’s GPU.

02

A hive you can keep.

FlorenciaAI is a self-hosted agentic assistant, and the beginning of a private operating system for a fleet of off-the-shelf humanoid robots — a hive mind that stays on your side of the wall. The same private-server technology is the basis for an on-premise offering for professionals who cannot put client data in the cloud. It is not a cloud service, and it is not generally available as a commercial product today. The founder’s private deployment is already running, and we are now accepting pilot inquiries.

Assistant

Agentic, local

Tool use, long-term memory, email, research, browser automation, and bot orchestration — designed to run entirely on owned hardware.

Offline

Open-weight only

The product runtime is local open-weight models. There is no dependence on a cloud frontier model for the assistant to think or act.

Later

A private OS for a robot fleet

FlorenciaAI will not build the robots. It will run and coordinate off-the-shelf humanoid robots from a private server, employed around town for everyday tasks. That stage is the long-term goal, not shipping today.

03

What’s running today.

A dedicated workstation named FloraAI, running Windows 11, already serves a custom 70-billion-parameter open-weight model on two NVIDIA GPUs, with a local interface, image and video tools, and a daily on-premises refresh pipeline.

Terminal rendering of a real offline FloraAI run, in which the model explains the FlorenciaAI robot-fleet operating-system vision, with verbose generation stats.
Real offline FloraAI output (Llama 3.3 70B, local): 17.2 tokens/s
Hardware
Two NVIDIA GeForce RTX 3090 Ti GPUs, 24 GB each (48 GB of VRAM in total), driver 591.86. Intel Core i5-13600K, 64 GB of RAM. The 70B model is split across both cards.
Primary model
Custom FloraAI, built on Meta Llama 3.3 70B Instruct (70.6 billion parameters), Q4_K_M quantization, 32K context, with a custom FloraAI system prompt for an agentic local assistant. Also a 64K-context variant (IQ4_XS). The base llama3.3:70b is Q4_K_M, served as GGUF by Ollama 0.40.16.
Runtime and interface
Ollama 0.40.16 on Windows 11. Open WebUI on the local network and on a phone.
Generation
ComfyUI on the same GPUs for image and video, including image-to-video and text-to-video tools.
Agents
Custom Ollama Modelfiles with agentic tools, including web search and ComfyUI-backed generation.
Measured inference
About 17.2 tokens/s generation on Llama 3.3 70B Q4_K_M (two runs, 17.29 and 17.19), measured on the FloraAI PC on November 18, 2026, at 9:52 PM ET. Cold prompt processing about 995 tokens/s on a 5,004-token prompt; 21.5 s cold load; 100% on GPU with no CPU offload, about 45 GB across both 3090 Ti cards.

FloraAI-Business

A local RAG business assistant, one of the custom FloraAI model builds, refreshed by a daily automated pipeline.

Daily pipeline

A Kasa smart plug wakes the FloraAI PC. The machine runs its refresh and backup, then shuts down.

Also installed

gemma3 12B and 4B, qwen2.5vl 7B for local vision, and nomic-embed-text for embeddings — so local vision and retrieval are possible.

This is a private, working prototype — not a public product launch. The generation figures above are from a cold-and-run measurement on the FloraAI PC, not a cached prompt-eval rate.

04

Private AI for professionals.

The first use of FlorenciaAI beyond the founder’s own workstation is a private, on-premise AI server for practices that cannot put client data in the cloud. Open-weight models run entirely on site, fully offline. Client files, prompts, and answers stay in the office.

This is the same private-server technology as the long-term vision: a central brain for a fleet of off-the-shelf humanoid robots. It is also FlorenciaAI’s first planned revenue line. We are not commercially shipping yet. We are now accepting pilot inquiries.

Law firms

Client matters stay in the firm.

Drafting, research, and review can run on a server in the office, so privileged work does not have to leave for a cloud model.

Medical practices

Patient context stays on site.

Notes, letters, and internal research can be assisted locally, designed to help meet confidentiality obligations without sending records to a public cloud.

Accountants

Financial files stay in the practice.

Returns, workpapers, and client questions can be handled on a machine the firm owns, with no dependence on a cloud frontier model.

Why on-premises

The model comes to the office. The data does not leave it.

Fully offline

Inference runs on hardware in the practice. An internet outage does not take the assistant with it, and a vendor’s cloud is not in the loop.

Open-weight models

The runtime is local open-weight models. There is no dependence on a cloud frontier model for the system to think or act.

Confidentiality by design

Client and patient data is meant to remain on premises. The design is meant to help meet confidentiality obligations. No industry certification is claimed.

How it works
  1. 01 A dedicated server stays in the office

    The system is installed on premises. FlorenciaAI provides the private AI software and the on-site deployment, not a hosted cloud workspace.

  2. 02 Models run entirely locally

    Open-weight language models run on that hardware. Prompts, documents, and answers are processed in the office.

  3. 03 The practice uses it on the local network

    Staff reach the assistant from machines on the premises. Client data is not sent to a frontier model operated by someone else.

  4. 04 Built on a working prototype

    The founder’s FloraAI workstation already runs a 70-billion-parameter open-weight model locally on two GPUs. That private server is the reference design.

Pilot program

Now accepting pilot inquiries.

If your practice needs a private AI server that never sends client data to the cloud, write to the founder. This is a pilot program around a working prototype, not a general product launch. There is no public price list, and no customers to list yet.

Inquire about a pilot jensen@florenciaai.com

Designed to help meet confidentiality obligations. FlorenciaAI is not a certified compliance product.

05

How the system is arranged.

Current private deployment. Planned serving and robotics paths are on the roadmap, not in this diagram.

06

Where it goes next.

Phase 1 is FlorenciaAI’s first planned revenue line: private, on-premise AI servers for professionals, now open for pilot inquiries. In parallel, the first robotics milestone is to have FlorenciaAI on our own server control a single robot arm within six months. After that, the path is to deepen the local hive, give it better serving and more capable on-premises NVIDIA hardware, and grow that private server into the operating system and central brain for a fleet of off-the-shelf humanoid robots. FlorenciaAI will not manufacture the robots. Nothing here is a purchase commitment.

Phase 1

Private AI for professionals

On-premise AI servers for law firms, medical practices, and accountants, running open-weight models fully offline, with no client data leaving the office. Now accepting pilot inquiries. Not commercially shipping.

Next 6 months

One robot arm, on our own server

FlorenciaAI on our own server controls a single robot arm — perception with the local vision model, planning with the 70B model — as the first step toward the fleet OS.

Planned

A stronger local hive

Richer memory, email, research, browser automation, and bot orchestration — still fully on-premises, still open-weight.

Planned

NVLink and more VRAM

Relieve the current two-card, x8, no-NVLink split so larger context, higher precision, and concurrent tools have room to run.

Planned

RTX PRO 6000 Blackwell

A 96 GB enterprise GPU, used only as internal FlorenciaAI development and deployment hardware if it is acquired.

Planned

TensorRT-LLM, vLLM, and NIM

Move serving from GGUF via Ollama toward NVIDIA TensorRT-LLM, vLLM, and self-hosted NIM on owned GPUs — not hosted cloud NIM as a runtime dependency.

Planned

A private OS for a robot fleet

Grow today’s private workstation into the central brain that coordinates off-the-shelf, hardware-agnostic humanoid robots doing everyday tasks around town. FlorenciaAI supplies the software and the compute, not the robots. Hardware scale comes later.

07

For investors.

FlorenciaAI is founder-led and bootstrapped, with $0 outside funding. It is pre-revenue and not commercially shipping. In November 2026, FlorenciaAI was accepted into NVIDIA Inception, NVIDIA's program for startups. We are now accepting pilot inquiries for private, on-premise AI servers.

Stage
Prototype · pilots open
Program
NVIDIA Inception member (Nov 2026)
Funding
Bootstrapped · $0 outside
Revenue
Pre-revenue
Shipping
Not commercially shipping
The opportunity

Phase 1, then the fleet OS.

The first planned revenue line is private, on-premise AI for professionals — law firms, medical practices, and accountants — who cannot put client data in the cloud. Open-weight models run entirely on site, fully offline.

The long-term vision is to grow that same private-server technology into the operating system and central brain for a fleet of off-the-shelf humanoid robots. FlorenciaAI supplies the software and the compute, not the robots.

What exists today

A working local prototype.

The founder’s FloraAI workstation already runs a 70-billion-parameter open-weight model locally on two NVIDIA RTX 3090 Ti GPUs, fully offline, with a local interface, agent tools, and a daily on-premises pipeline.

See what’s running for the measured setup. That private server is the reference design for the professional offering and for the later fleet OS.

Near-term milestones
  1. Pilot inquiries for professional servers Phase 1 revenue, now open. Not a general product launch.
  2. One robot arm, on our own server Within six months: local vision for perception, the 70B model for planning.
  3. A stronger local hive and better serving Richer on-premises tools, then TensorRT-LLM, vLLM, and self-hosted NIM on owned GPUs.

Investor materials

Request the investor deck.

Write to the founder for the current deck and a conversation about FlorenciaAI. This page is informational only.

Request the investor deck jensen@florenciaai.com

This is not an offer to sell securities.

08

Who is building it.

CEO & Managing Member

Tut Liu

Tut Liu is a Worcester, Massachusetts entrepreneur, artist and marketer with more than 20 years of creative experience. He has run Tut Liu Photography since 2003, building deep skills in photography, graphic design and brand storytelling.

Tut is a co-founder and the Chief Marketing Officer of Delivered, Inc., a licensed cannabis home-delivery company serving Central Massachusetts, which launched in July 2023 as the fifth licensed home delivery operator in the state. He leads its marketing and community engagement, including coat drives, continuing-education scholarships and a planned free incubator program for budding entrepreneurs.

Tut comes from an immigrant background, and giving back to the community is central to how he works. As CEO and Managing Member, he brings experience in regulated-industry operations, brand building and community-focused growth.

Founder

Jensen Florencia

Jensen Florencia is a Massachusetts serial entrepreneur with a background in financial services, where he previously held FINRA Series 7, 24, 63 and 65 licenses. Years in that highly regulated industry gave him a deep respect for compliance and for protecting client data, which is the problem FlorenciaAI's private servers solve.

Jensen holds an MBA from Boston University and a Bachelor of Science in Business Administration from Worcester State University, and he is fluent in English and Spanish.

He builds and runs the FloraAI private-server prototype, which already serves a 70-billion-parameter open-weight model on two RTX 3090 Ti GPUs.

Contact

Write to the founder

Pilot inquiries, investor inquiries, and anything else about FlorenciaAI go to the same address.