Technology portfolio · v4.1 · 2026

We build full-stack AIoT systems — and we ship our own.

Hardware, firmware, edge AI, connectivity, cloud and mobile for robotics, mobility, energy and smart consumer electronics. Six engineers, deliberately one deep per layer, across Singapore and Ho Chi Minh City.

Robotics
Mobility & EV
Energy
Smart consumer
Industrial IoT

Download the PDF [email protected]+65 9085 2748

34-page PDF · 5.4 MB · same content, laid out for a meeting room

13+: years of embedded engineering leadership
6: engineers, deliberately one deep per layer
20+: hardware and firmware programmes delivered
5: own products in market or in build

Capabilities

Six layers, one supplier.

Engage us for the whole stack, or drop us into the layer where your team is thin. Every layer has a named owner who has shipped that layer before.

Hardware

Sensors, cameras, actuators, MCUs and high-speed processors. Schematic, PCB layout and DFM in-house, then bring-up, certification and supplier management.

Schematic & PCB layout
DFM
Bring-up
Certification

Firmware & embedded

Zephyr, FreeRTOS, OSEK and embedded Linux built with Yocto. AUTOSAR and ADAS experience. Rust where the firmware has to be safe and secure by construction.

Zephyr RTOS
FreeRTOS
Yocto Linux
AUTOSAR
Embedded Rust

Edge AI

TinyML and quantised int8 vision at low power, on-device LLM integration, and the training and evaluation tooling that proves a model before it ships.

TinyML
Quantised vision (int8)
On-device LLM
OTA models

Connectivity

MQTT, Zenoh, Wi-Fi, BLE and BLE Mesh, LoRa, NFC, LTE and CAN — chosen against the range, latency and milliamp budget the product actually has.

CAN / CAN-FD
BLE & BLE Mesh
LoRa
Wi-Fi 6
LTE
Zenoh

Cloud & backend

Scalable, real-time services with time-series storage, processing and ML model integration. Rust micro-services. Data stays where the regulator expects it.

Rust services
Time-series storage
MQTT broker
Model serving

App & interface

Flutter and native mobile, plus desktop tooling in Rust. The interface is specified against the same latency budget as the firmware under it, so what ships behaves like what was designed.

Flutter / Dart
Rust + Slint
Embedded Flutter
Design system

What that means for you. A small team with no gaps — so there is no integration seam where three vendors meet, and no discipline waiting on a hire. Around all six layers we also run certification, first-batch planning and supplier management, because we have taken our own products through it.

Technology stack

What we actually build with.

No exploratory learning on your budget — this is the stack we run every day.

Silicon: NXP i.MXRT
NXP i.MX8
NXP S32K
ST STM32N6
ST STM32U0
Nordic nRF52840
Espressif ESP32
Operating systems: Zephyr RTOS
FreeRTOS
OSEK
Embedded Linux (Yocto)
Android Automotive
Languages & AI: Rust
Embedded Rust
C / C++
TinyML
Quantised vision (int8)
On-device LLM
Connectivity & cloud: CAN / CAN-FD
BLE & BLE Mesh
LoRa
MQTT
Zenoh
Wi-Fi 6
LTE
Time-series cloud

PCB design in-house · Rust desktop tooling · Rust backend micro-services

Part one

Products we own.

We do not only build for clients. These are ours — designed, engineered, manufactured and sold by BluLeap. A client is buying a team that has already carried its own product risk.

Our product · veeaai.com

The little camera that tells you what happened

Tell Veea what matters in your own words. It watches quietly, sends you a sentence the moment it happens, and deletes the photo seconds later.

Say it, don't configure it. No motion zones, no rule builders. “Tell me if the stove is left on.” “Watch the door while we're away.” Plain language in, a sentence out.
A diary, not a video feed. Every event becomes text. Ask when the plumber came and get an answer instead of an hour of scrubbing — plus daily summaries for helpers and elderly parents.
Privacy in the silicon, not the settings page. A European secure chip reads every frame on the device, only event snapshots reach the cloud, and images of people are never stored. A tamper-proof identity chip per camera.
A light that never lies. Off means private. A slow coral breath means watching. A double pulse means Veea messaged you — verifiable from across the room, no app required.

On-device inference
Full HD · 160° lens
Wi-Fi 6
USB-C or 4×AAA
IP67
Secure element

Working prototype · 38 × 38 mm · first batch early 2027

veeaai.com

The Veea camera prototype standing on a desk — Printed and assembled 66 mm shell, running the vision pipeline

The Veea camera held in one hand — Held in one hand — the size the industrial design was specified around

Front of the Veea camera showing the lens barrel and status light — Front shell, lens barrel and status light, as assembled

Veea app showing a written summary of the day — A day of activity, condensed into a paragraph

Veea app showing events listed by date — Pick a day and read what happened — no timeline to scrub

Veea app showing a safety alert written as a sentence — A possible fall surfaces as a sentence with one action attached

Veea app screen for writing a rule in plain language — Tell it what to watch for, in your own words

Veea app searching the camera memory in natural language — Ask the camera memory a question, across days

Veea app showing a rule being applied to a live scene — The user sees the rule being applied

Our product · open source

Four keys for the agent that keeps asking

AI coding agents stop and ask. Keycap Context turns the question into a native macOS overlay and four physical keys, so the answer is a keypress instead of a context switch.

One overlay, every agent. Claude Code, Codex, Gemini CLI, Antigravity, OpenCode and a generic JSON adapter all render into the same overlay, with the agent’s own wording and ordering preserved. Keys one to four light up as the visible choices.
Gestures carry the rest. Short press selects. Long press pages, clears a multi-select, hands the request back, or confirms a destructive action. Double press moves between pages. The firmware holds a short press for 280 ms so it can tell one from a double.
Four agents on four keys. Agent Console assigns recent sessions to the keys and colours them by state — pulsing blue working, orange waiting, pink for a decision that can delete something, green done. Press to focus that terminal, hold to interrupt it.
Local-first, and it proves it. The host listens on loopback with a per-install token and needs no macOS Accessibility permission. The audio-reactive standby effects run on the keypad’s own microphone: sound becomes four band levels on the device, and neither the audio nor anything derived from it crosses the USB link.

Zephyr RTOS
XIAO nRF54L15 Sense
NeoKey 1x4
Swift / AppKit
Local HTTP API
Apache-2.0

Open source · firmware, macOS host, adapters and CAD

github.com/0xkelvin/Keycap-Context

Audio-reactive standby — the microphone drives the LEDs on the device itself

A hand holding the four-key pad in front of a monitor, where the Keycap Context overlay is showing three numbered choices from Claude — The agent asks on screen; the keys light with the choices

The Keycap Context overlay showing a question from Claude with four numbered choices and a WAITING badge — The overlay, built from the agent’s own choice list

Keycap Context settings window with lighting effect, brightness, speed, per-button colours and gesture assignments — Lighting, gestures and context profiles, in the menu-bar settings

The four-key NeoKey pad lit blue, sitting on a laptop and wired back to the XIAO nRF54L15 board — NeoKey 1x4 and a XIAO nRF54L15 Sense, wired on the bench

Our product · eng.veeaai.com

Vocabulary that follows your real work

A personal English companion for professionals: capture the words you actually meet at work, then keep them.

Capture where you read. Desktop OS-level capture, in-app entry, and a camera lens that breaks a page into tappable words — so a word goes in at the moment you meet it, not in a study session later.
AI enrichment. Synonyms, antonyms, usage in context and memory images generated per word, so a word arrives already learnable.
Review without a screen. The day's words play as a tape on the commute, with a recall pause, a repeat count and a word-only or word-with-example voice.
An arcade, not a quiz. The same list drives Snake, Invaders and a sixty-second rush, and a link code drops a word duel on a friend.
E-ink-native, so it surfaces anywhere. High contrast, monospaced type and no reliance on colour, which is why the same design works on the lock screen, the home widget and the 1.54" e-ink keychain — a Nordic nRF52 syncing over BLE to the phone only. A word goes in once and keeps coming back until it sticks.

Flutter / Dart
Rust backend
SQLite
OCR capture
Text-to-speech
BLE GATT
Spaced repetition
On-device

Live · free, pro and hardware-bundle tiers

eng.veeaai.com

The Veea journal for a day, listing each word with its phonetic spelling, its Vietnamese meaning and a speaker button — The day's journal — every word with its sound and its meaning

Pixel Lens OCR screen showing a captured page broken into individual tappable words, with the selected word ready to save — Pixel Lens — snap a page, tap the word you did not know

Commute player styled as a cassette deck, playing the day’s words with recall pause, repeat count and voice mode settings — Commute player — the day's words as a tape, with a recall pause you set

Arcade centre listing vocabulary games including Pixel Link Duels, Word Rush 60s, Vocab Snake and Vocab Invaders — The arcade — one word list, five cartridges

Vocab Snake game, steering a snake towards the word pellet that matches the meaning shown at the top — Vocab Snake — steer to the pellet that matches the meaning

Vocab Invaders game, firing at alien ships labelled with English words to match the Vietnamese prompt at the top — Vocab Invaders — shoot the ship carrying the right meaning

Pixel Link screen showing a personal game link code, a field to connect with a friend, and connected friends with word drop buttons — Pixel Link — a code connects two players and drops a word on a friend

Veea English word of the day widget on an iOS home screen — Word of the day, on the home screen

A phone lock screen showing the day's word with its phonetic spelling, its Vietnamese meaning and a streak counter — And on the lock screen, before the phone is even unlocked

Our product · tracking

An e-ink tag on both finding networks

A Zephyr-based e-ink tag: custom GATT services, an on-phone image pipeline, and both major finding networks.

Custom GATT for reliable sync. Purpose-built GATT services for high-reliability mobile sync, tuned for intermittent connections rather than lab conditions.
Image pipeline on the phone. Dither and pack a photo to the panel's native format on-device, then stream it in chunks with a commit handshake.
Apple Find My and Google Find Hub. Third-party tracking protocol integration on both ecosystems, meeting their security and performance compliance bars.

Zephyr RTOS
BLE GATT
Apple Find My
Google Find Hub
E-ink
Seeed XIAO nRF52840

Open source · github.com/bluleap-ai/the-tag

github.com/bluleap-ai/the-tag

Phone to panel over BLE — 5,776 bytes in 24 chunks, a commit, then the refresh

E-ink tracking tag concept next to a phone — Industrial design concept — target form factor

Working e-ink driver board with a dithered image on the panel — Working driver board and 24-pin flex to the panel

Our product · wearable

An AI that remembers for you

An always-on memory capture pendant. Capture is the easy half — the product is what comes back: a structured record and a task list you did not type.

Structured, not a transcript wall. A conversation is broken into topics with timestamps — reviewable in seconds instead of replayed in full.
Commitments become tasks. The system pulls out what was agreed and files it as a dated reminder, so the follow-through does not depend on memory.
Built end to end. Enclosure, board, firmware, BLE link, backend and mobile app — all designed in-house, which is what makes the loop this short.

Always-on audio
BLE
Low-power MCU
Cloud transcription
On-device wake control
Multi-day battery

Prototype · one body, four ways to wear it

The Veea wearable worn as a pendant — Worn prototype — pendant mode

Industrial design study showing colourways and wear modes — Industrial design study — colourways and wear modes

App screen showing a meeting segmented into timestamped topics — Meetings, segmented and timestamped

App screen showing action items extracted from a conversation — Action items, extracted automatically

Our product · safety

Smell the fire before there is smoke

Conventional alarms wait for smoke and heat. We train on the odour of overheating plastic and wire — the signal that arrives first.

Sense the odour, not the smoke. Custom gas sensors sample the air in high-risk spaces — electrical cabinets, EV charging bays, warehouses.
Classify at the edge. Inference runs on-device; nodes mesh to each other and to a local AI hub over Bluetooth mesh, with OTA model updates.
Retrofit, don't replace. Alerts feed people and existing fire alarm installations, so a building does not have to be re-certified to get the warning.

Custom gas sensing
On-device classification
Bluetooth mesh
OTA models
Training tooling

Top 16, AI Việt Innovation 2024

Live bench classification of a heating plastic sample — Live bench classification of a heating-plastic sample

Model training tool showing labels, training status and a confusion matrix — Training tool — label, train, inspect the confusion matrix

Part two

Built for clients.

Selected programmes across robotics, mobility, energy, industrial and consumer hardware.

Consumer · health & motion

Wearable AI for gait and motion tracking

Hardware, firmware and on-device AI for a bio-motion system — one sensor stack serving clinical gait metrics and sports performance.

Hardware. Compact, low-power board with accelerometers, gyros and health sensors in an ultra-light form factor, strapped where the gait signal actually is.
Firmware. Optimised C/C++ on ARM MCU with a BLE stack for phone sync, real-time logging, power management and OTA.
Edge AI. On-device motion classification deriving cadence, step length, foot clearance and speed from raw IMU — not just a step count.

Nordic nRF52840
Bosch 9-axis IMU
BLE
OTA
Edge inference
C / C++

Counting on the walk — steps, cadence, foot clearance and speed, live off the ankle sensor

App showing live gait metrics including cadence and foot clearance — Gait metrics, live

App showing a sports performance view with match statistics — Sports performance view

Ankle-mounted gait sensor prototype strapped above the ankle — Ankle-mounted prototype

Robotics · mobility

Open-source telematics platform

Our own tracker hardware and cloud stack, published as open source — deployed on fleet and robotic units.

Board we designed. ESP32 with a cellular modem, GNSS, dual SMA antennas, USB-C and a screw-terminal vehicle harness — sized to sit on a bike, not a bench.
Real-time CAN to cloud. CAN data and GNSS position forwarded live over MQTT, with OTA firmware delivery to ECUs across the CAN network, rollback included.
Trip-level telemetry. Every ride resolved into a replayable route with duration, distance and state of charge — plus crash and fall detection with immediate alerts.

ESP32
LTE
GNSS
CAN / CAN-FD
MQTT
OTA over CAN

github.com/bluleap-ai/telematic-esp32

Telematics tracker board with ESP32, cellular modem and dual SMA antennas — Tracker board — ESP32, cellular and GNSS

Dashboard showing a replayed trip route and per-ride telemetry — Trip replay and per-ride telemetry in the dashboard

Energy · industrial

Transformer fault detector with integrated gateway

Detect and re-energise failed distribution transformers remotely, without sending a truck out first.

Remote fault detection. Detect a failed transformer and re-energise it remotely, cutting the response loop from a truck roll to a command.
Two boards, one enclosure. A dedicated ADC daughterboard carries the analog front end; the main board handles compute, isolated rails and backhaul.
Condition monitoring at the edge. Continuous condition data with remote diagnostics over MQTT, on an NXP i.MXRT dual-core running Zephyr.

NXP i.MXRT dual-core
Zephyr RTOS
Isolated DC-DC rails
Cellular + Wi-Fi/BLE
MQTT
Sealed field connectors

Transformer fault detector main board and ADC daughterboard — DVT hardware — main board and ADC daughterboard

Rendering of the detector module mounted on a distribution transformer — Deployment: module mounted on the transformer

Mobility · safety

ADAS box for mass-market vehicles

Advanced driver assistance for budget cars, taxis and buses — features that used to be reserved for premium trims.

Edge vision, no vehicle surgery. On-device object detection over a dash camera feed, designed for rapid deployment without touching the vehicle's core systems.
The safety set. Overspeed detection and warning, forward collision warning, and detection of vehicles, pedestrians and obstacles.
Real-time overlay. Alerts rendered live over the camera feed so the driver sees the reason for the warning, not just the warning.

Edge object detection
Computer vision
Real-time overlay
Retrofit

Shipped 2020; the vision stack still underpins our edge AI work

Automotive · HMI

In-vehicle HMI on custom Linux

Hardware and software for an automotive human-machine interface — from a display on the bench to a cluster in the dash.

No black box in the boot path. NXP i.MX8x on a custom Linux built with Yocto — a board support package we build and maintain ourselves, boot and device tree included.
Modern UI, automotive base. Toyota Connected's Flutter IVI homescreen and Sony's embedded Flutter for Linux, kept upstream-aligned rather than forked into a corner.
Hardware and UI specified together. Display, touch and compute chosen against one latency target, so the interface that ships behaves like the one designed.

NXP i.MX8x
Yocto Linux
Embedded Flutter
Android Automotive
Custom BSP

HMI bring-up rig with panel, carrier board and bench supply — Bring-up rig — panel, carrier board and bench supply

Instrument cluster UI running inside the enclosure — Cluster UI running in the enclosure

Automotive · boot time

Three seconds from power-on to a live cluster

Client work on boot time for an HMI cluster: cold start to a fully rendered instrument cluster on NXP i.MX 95, timed on a stopwatch sitting in the same frame as the screen.

Measured, not quoted. The stopwatch runs in shot beside the display and the run is repeated. What you see is the number the platform holds, not a best case pulled out of a boot log.
The whole path is the budget. Bootloader, kernel, device tree, display bring-up and the cluster's own first paint all have to fit inside the three seconds. The work is trimming every stage, not just the last one.
A cluster, not a splash screen. Speed, range, state of charge, ride mode, temperature, power and trip data are all on the screen the moment it lights — there is no placeholder frame standing in while the rest loads.

NXP i.MX 95
Embedded Linux
Boot-time optimisation
HDMI
EV instrument cluster

Measured on an NXP FRDM i.MX 95 development platform, driving the cluster over HDMI

Cold start to a rendered cluster — the stopwatch reads 2.95 s

EV instrument cluster on the monitor showing 188 km/h, 249 km range, 100% charge, ride mode and trip data — First paint — speed, range, charge, ride mode and trip data all up at once

Consumer · gaming peripherals

8 kHz analog input, measured not assumed

Firmware for high-polling-rate analog-switch keyboards — and the two measurement rigs that prove the numbers.

Analog signal chain. Per-key ADC acquisition with individual calibration curves, drift compensation and hysteresis — travel resolved to microns, not a binary press.
Deterministic 8 kHz path. A 125 µs report budget end to end: scan, filter, decide, report. The work is in bounding jitter, not hitting the average.
Two measurement paths. Our own Rust and Slint tool for per-key ADC and travel; the open-source XLAT rig on an STM32F7 for end-to-end click-to-photon latency. Firmware-side numbers and system-level numbers, cross-checked.
Bring-up to production line. First power-on and rail verification, an on-device GUI inside the input latency budget, 2.4 GHz and Bluetooth variants, and signed, encrypted OTA through a WebAssembly web app — plus test builds that let the factory screen boards.

USB HID
8 kHz polling
Analog / TMR switches
Per-key ADC
NKRO
Rust
Slint
XLAT on STM32F7

Per-key switch and LED matrix with an integrated display cut-out

Characterisation tool showing per-key ADC values and calibration — Our characterisation tool — per-key ADC and calibration

XLAT click-to-photon latency measurement rig — XLAT click-to-photon rig

Consumer · open source

Open-source keypad — our board, our firmware

A keypad platform we designed and built ourselves: schematic, layout, plate fit and firmware, published rather than filed away.

Hardware we designed. Controller board with USB-C, key matrix and per-key LED footprints, a rotary encoder in the matrix, plate mounting and 2U stabiliser fit.
Firmware we wrote. USB HID enumeration, matrix scan and debounce, encoder decoding and key remapping — the same pipeline as our commercial input work, at a size a client can read in an afternoon.
Open by intent. Publishing the design means a prospective client can inspect our schematic, our layout and our source before signing anything. That is a faster trust route than any slide.

USB-C
USB HID
Key matrix scan
Rotary encoder
Per-key LEDs
Hot-swap sockets
Open hardware

Assembled open-source keypad prototype with board, plate and encoder — Assembled prototype — board, plate, stabilisers, encoder

Energy · metering

Smart meter, bring-up to production firmware

Hardware bring-up, firmware implementation and purpose-built test firmware for a utility-grade smart meter programme.

Hardware bring-up. First power-on, rail verification and probing the board back to schematic — the phase where a meter programme either finds its problems or ships them.
Firmware implementation. Metering firmware on the target: measurement path, LCD driving, push-button UI and WAN communication over the cellular link.
Firmware built for test. Purpose-built builds that exercise each subsystem so the line can screen boards — unglamorous work, and the reason volumes hold their yield.

3P4W HTCT
Class 0.5S
IS 15959 Part 3
4G / 2G WAN
LCD & keypad UI
Test firmware

Production three-phase four-wire smart meter with the branding on its rating plate obscured — Production meter — 3P4W HTCT, Class 0.5S

Assembled smart meter with its board, LCD and push buttons visible through the sealed transparent cover — Assembled unit — board, LCD and push-button UI behind the sealed cover

Bench setup with a smart meter board wired to a portable three-phase test set, alongside a laptop and a multimeter — Accuracy checks — the meter wired to a portable three-phase test set

Smart meter board on the bench with rails probed back to schematic — Bring-up on the bench — rails probed back to schematic

Consumer · pet tech

Connected dog collar

Hardware and firmware for a pet wearable: live position, step count and temperature, on a battery that has to survive a week on an animal.

The board we built. Quectel cellular module, u-blox NORA-B106 for BLE, u.FL antenna feeds, USB-C with LiPo charge management, and an FPC link to the sensor board.
Power is the product. Duty-cycled position fixes and cellular reporting. The difference between a collar worn every day and one that lives on the charger is milliamps, not features.
Behaviour, not raw traces. Steps, temperature and position rolled into daily, weekly and monthly views — plus a movement heatmap an owner can read at a glance.

Quectel cellular
u-blox NORA-B106
GNSS
LiPo charge management
FPC sensor link

Dog collar mainboard with cellular, BLE and charge management — Collar mainboard — cellular, BLE and charge management

The connected collar in use on a dog, with the companion app — The product in use

Sensing · environment

Hyperlocal environmental sensing

Compact nodes that turn a city into a live air, light and noise map — one street corner at a time.

Two MVP generations. A modular v1 for sensor selection, then a compact round v2 with the radio, battery management and sensing on one board.
Nine signals per node. PM2.5, PM10, NO₂ and ozone, UV index, noise level in dBA, humidity, temperature and gas resistance — all on a duty-cycled power budget.
Density is the product. Readings are hex-binned into a live map, so a resident sees their own street instead of a citywide average.

ESP32 Wi-Fi/BLE
PM2.5 / PM10 / NO₂ / O₃
UV index
Noise level (dBA)
Mobile dashboard

The MVP v1 sensing board: an ESP32 module with Air Quality 8, Air Quality 11 and Environment 3 click boards, a UV 5 slot and a 2000 mAh lithium cell alongside — MVP v1 — a modular board for choosing the sensors

Compact round Wi-Fi and BLE environmental sensing node held in a hand — MVP v2 — radio, battery management and sensing on one board

Mobile app showing a hex-binned city air quality map — Hex-binned city map

Mobile dashboard showing live environmental metrics — Live metric dashboard

R&D · wearable platforms

Open-source smart glasses platform

Hardware and firmware for an open-source eyewear platform — electronics that have to disappear into a few millimetres of frame.

Everything inside a temple arm. Flex PCB, battery, speaker, mic and optical sensing packed into an arm that still has to look like eyewear and sit on a face all day.
An open platform, not a locked device. Hardware and firmware published for the curious and the creative, so third parties can build on the platform rather than around it.
Comfortable at TRL 3–5. We take early-stage platform work, not only production programmes — and what we learn at that stage feeds straight back into the production stack.

Flex PCB
Optical sensing
Audio
BLE
Open hardware

Smart glasses prototypes opened on an ESD mat, showing the flex board, speaker and antenna that live inside the temple arm — Opened on the mat — flex board, speaker and antenna from one temple arm

Two smart glasses prototypes wired out to breakout boards and a USB programmer for flashing and debug — Two units broken out to a programmer for flashing and debug

Workbench with several smart glasses prototypes on a mat, a bench power supply, a multimeter and debug probes — The bench — prototypes on the mat, with bench supply, meter and debug probes

The platform

What every device we ship gets for free.

The fleet layer is built once and reused — you do not pay to reinvent it per programme. Already deployed in production across telematics, energy and consumer programmes.

Over-the-air updates

Secure firmware upgrade for devices and ECUs
Delivered via cloud or local server
Rollback and version control

Remote diagnostics

Real-time health of deployed devices
Detect sensor failure, comms loss, overheating
Metrics visualised in a cloud dashboard

Remote services

Parameter tuning: sampling rate, thresholds, triggers
Server-initiated commands and status requests
Predictive maintenance and support

From embedded sensor to early warning

The same pipeline behind our transformer monitoring, telematics fleets and Veea's event diary.

01
Ingest
Data from embedded sensors, buffered on device and forwarded when the link allows
02
Store
Time-series storage plus retained historical data for training and comparison
03
Transform
Cleaning, feature extraction and aggregation into analysable series
04
Analyse
AI and predictive algorithms run over live and historical data
05
Deploy
Real-time alerts and early warnings into OEM, dealer and operator applications

Predictive maintenance is the commercial outcome: a failure detected two weeks early is a service call instead of a site visit, a recall, or an outage.

Why BluLeap

Embedded intelligence, not disconnected parts.

Most suppliers build one layer and integrate badly at the seams. We build all six — and we prove it on products of our own.

We carry our own product risk

Veea is our camera: our silicon choice, our industrial design, our app, our cloud, our certification and our first batch. A client is buying a team that has already been through it, not one that has only advised on it.

One team from board to backend

Proprietary VCU and telematics stack, embedded AI, smart sensor integration and Flutter UI/UX — under one roof, so there is no integration gap where three vendors meet.

Two-country execution

Singapore for standards, market access and client proximity. Ho Chi Minh City for engineering depth at a cost base that lets a programme run longer on the same budget.

How to engage us

Full-stack product development
Concept to production, with one team accountable for every layer.
Custom AI integration
Intelligence at the edge of hardware you already ship.
Rapid prototyping
An MVP built fast, then hardened into something manufacturable.
IoT & embedded consulting
Deep-tech guidance, architecture reviews and firmware audits.

What clients get out of it

Real-time remote visibility into product usage and performance
Longer product lifetime through predictive diagnostics
Faster time to market via prototyping and modular systems
Competitive differentiation through customised hardware intelligence
Lower operational cost — up to 40% in deployments we have measured

Let's build

Tell us which layer you are missing.

Send the problem, not a brief. We will tell you what we would build, who on the team owns it, and roughly what it costs.

Singapore: SingaporeStandards, market access and client proximity
Ho Chi Minh City: Floor 14, HM Tower, 412 Lý Chính Thắng, District 3, Ho Chi Minh City, Vietnam 700000Engineering depth — hardware, firmware, AI, cloud

Our products · veeaai.com · eng.veeaai.com