ProventusNova
Let's get started

Ready to kickstart your
EdgeAI project?

Watch the 5-minute video below. Then book a 30-minute discovery call. We'll tell you exactly what's blocking your hardware and whether we're the right fit.

Technologies We Work With

NVIDIA
MediaTek
GStreamer
Yocto Project
NVIDIA
MediaTek
GStreamer
Yocto Project
Step 2 · Book your call

Pick a time. Andrés takes every call himself.

30 minutes. No deck, no fluff. We talk about what's blocking you and what it would take to unblock it.

Selected Work

We've solved this before.

Five engagements. Five different platforms and product stages. Every one shipped.

01 / 05
Race Car AI on NVIDIA Jetson
Robotics · Edge AI · NVIDIA Jetson

Race Car AI on NVIDIA Jetson

ProventusNova shipped the embedded AI foundation for a remotely operated race car. Lane detection, collision avoidance, self-driving home. All on NVIDIA Jetson at race-track latency.

Time to unblock the AI stack
2 weeks
Engagement length
8 weeks
Outcome
Follow-on round secured
02 / 05
UncommonLab USB Rescue
Autonomous Driving · NVIDIA Jetson AGX · JetPack 6

UncommonLab USB Rescue

ProventusNova cleared a weeks-long USB stack blocker on UncommonLab's custom NVIDIA Jetson AGX autonomous driving platform. We shipped the fix in 30 hours.

From engaged to resolved
30 hours
JetPack versions patched
6.2 + 5.1
Outcome
New contract on Jetson Thor
03 / 05
$200 Off the BOM. Same AI.
Healthcare · NVIDIA → MediaTek

$200 Off the BOM. Same AI.

ProventusNova migrated a healthcare device's machine-learning workload from NVIDIA Jetson Nano to MediaTek Genio 510. Cut per-unit cost without losing inference performance.

Saved per unit shipped
$200
Time to US-market launch
12 months
Outcome
US-market dominance
04 / 05
Off-Grid AI in Your Hand
Privacy-First · MediaTek Genio · On-Device AI

Off-Grid AI in Your Hand

ProventusNova is building a privacy-first off-grid handheld with on-device speech AI, encrypted mesh networking, and a Matter smart-home bridge. All running locally on MediaTek Genio with zero cloud calls.

AI inference
100% on-device
Encrypted mesh
Reticulum
Security
Pen-test stage-gate ready
05 / 05
ROSORIN, Reborn on Genio
Concept Build · Robotics · Edge AI

ROSORIN, Reborn on Genio

Three concepts ProventusNova would prototype on the Hiwonder ROSORIN platform. Developed against NVIDIA Jetson, shipped on MediaTek Genio.

Concepts ready to prototype
3
Flagship concept
On-device VLM assistant
Reference platform
Hiwonder ROSORIN
Testimonials

The teams who hire us
hire us again.

Bongjin Raum Jeong

We had been blocked for several weeks trying to get USB working on our custom carrier board with JetPack 6. ProventusNova resolved it in under a week.

Bongjin Raum Jeong

CEO, UncommonLab

Milan Young

“In a one-hour working session, they got the board booting in minutes. Saved us days of trial and error.”

Milan Young

CTO & Founder, Farmhand AI

Haneul Louis Yoon

“Critical USB fixes ported to JetPack 5 in just 10 hours. Kept the project on schedule.”

Haneul Louis Yoon

CFO, UncommonLab

Aiswarya Rani

“Deep expertise in embedded systems made our integration seamless. Delivered on time, exceptional quality.”

Aiswarya Rani

Engineer

How we work

What working with us
actually looks like.

01

Direct line to Andrés

You talk to the CTO who runs your project. No account managers, no offshore handoffs, no middlemen.

02

Daily Slack updates

A written log at the end of every working day. What we tried, what worked, what is next. You always know where the project stands.

03

Weekly working demo

A live screen-share at the end of every week. We show you the actual hardware running. No slide decks. No status theatre.

04

Full IP transfer

Every kernel config, driver, build script, and piece of documentation ships to your repo. Zero vendor lock-in.

The roadmap

14 days. 3 phases.
Hardware that runs.

Every Proof Sprint follows the same structure. You know exactly what happens and when.

Day 0

Qualification

30-minute call with Andrés. We confirm the engagement is in our wheelhouse and the scope is bounded. Go / no-go decision. No-go costs you nothing.

Days 1–3

System Foundation

Build environment, base BSP for your Jetson module or Genio chip, critical dependencies, early bootloader validation. The foundation everything else stands on.

Days 4–11

Execution Sprint

BSP, drivers, camera pipeline, AI inference, whatever your sprint is scoped to. Daily Slack updates, weekly demo. We hit the milestone or you do not pay.

Days 12–14

Handoff

Final stability validation, reproducible build documentation, full code handoff to your team, walkthrough session. You own the result forever.

Zero engagement risk

2-week trial. Pay nothing if it's not working.

After two weeks of work, if you're not completely satisfied (the quality, the communication, the progress), you pay nothing and keep everything we've built.

We can offer this because we only take on projects we already know how to deliver. We qualify every engagement before we start. A no-go decision costs you nothing.

FAQ

Common questions

What happens on the discovery call?
A 30-minute working session. We dig into where your hardware is blocked (board bring-up, camera pipeline, kernel issues, AI deployment) and tell you exactly what it would take to unblock it. If we're a fit, you walk away with a clear 14-day path. If we're not, you get a real diagnostic of what's blocking your project.
How fast can you start?
We take on 2–3 Proof Sprints per month and are typically booked about 2 weeks in advance. If your demo, pilot, or milestone is in the next 30–60 days, book a call this week.
How does pricing work?
Fixed-bid. You know the cost before we start. 50% upfront, 50% on delivery. Entry point is the Proof Sprint: 14 days, one bounded milestone, fixed price. No hourly surprises.
What if we're not satisfied after two weeks?
You keep everything we built and pay nothing. We only take on projects we already know how to deliver, which is why we can offer the guarantee.
What platforms and stacks do you cover?
NVIDIA Jetson (Orin AGX/NX/Nano, Xavier) and MediaTek Genio (510, 700, 1200). Board bring-up, BSPs, Yocto, kernel work, CSI/GMSL2 cameras, GStreamer pipelines, TensorRT, NeuroPilot. We don't do generic embedded work, we go deep on these two platforms.
Do we own the code afterwards?
Yes. Full IP transfer on completion. Every kernel config, driver, build script, and piece of documentation goes to your team. Zero vendor lock-in.
Andrés runs every discovery call?
Yes. Andrés Campos (CTO, co-founder) takes every call himself. You're talking directly to the person who would lead your project.

Book the call.
See if you qualify.

If we can't help you with your project, we'll tell you on the call and show you the best path forward.

Pick a time

Two Proof Sprints per month. Typically booked 2 weeks out.