AVA Light, three-quarter product render on white

ProblemBuilt for my sister Ava, who couldn't modify her rental's ceiling lighting. Warm LEDs still suppress melatonin, candles risk fire, smart bulbs need apps.

ApproachUse the existing white ceiling as a large diffuse reflector. Single fixed 1800K LED firing upward, fully analog dimming, passive thermal, USB-C PD.

ToolsFusion 360 for design, Bambu Lab 3D printers for housing iteration, overseas CNC for the aluminium heatsink, in-house rigs for electrical and thermal QA.

ResultSold-out first batch at A$489 retail, ~50% margin. Fully internal design, electronics, thermal, manufacture, and validation. No firmware, no recalls, no service overhead.

Overview

AVA is named after my sister. She rented a flat with fixed ceiling lighting she could not modify, and wanted something that supported her evening wind-down without the usual trade-offs: 2700K warm bulbs still suppress melatonin, candles introduce particulates and fire risk, and smart bulbs bring app dependencies, network setup, and firmware maintenance. There was no product on the market that did one thing well: low-CCT, glare-free, install-free, no-firmware lighting.

AVA is an upward-firing light. The LED projects 1800K amber light onto the ceiling, which acts as a large diffuser and washes the room in indirect glow. There is no visible LED from any horizontal viewing angle, so the source is glare-free. The light holds a stable 1800K (warm amber) output across the full 0 to 100 percent dimming range. Most off-the-shelf warm LEDs shift colour temperature as they dim and lose CRI; AVA does not move colour temperature at all.

Designed, engineered, manufactured, and sold end to end under my own brand, OUDESIGNLAB. Hand-assembled in Melbourne. Retails at A$489.

Visuals

AVA Light, full three-quarter portrait with knob detail

AVA Light at full output, ceiling-bounce glow on a wall

In situ

Ceiling shadow patterns cast by AVA Light through rattan pendants, the ceiling-bounce effect in action

Open-plan room lit by a single AVA Light from above, showing the spatial coverage

People around a table lit by AVA Light, evening atmosphere

Electronics & control

Fully analog. A potentiometer (PTR90) with a fully CNC-machined aluminium knob sets the dimming reference for a constant-current driver (MEAN WELL LDH-25-700). Input is USB-C PD at 20 V via an INJOINIC IP2721 PD trigger, so any 30 W+ PD charger powers the unit. No microcontroller, no PWM, no firmware, no app. The output is flicker-free.

Optics & thermal

The LED is a Luminus CTM-14 chip-on-board, 1800K, 90+ CRI, driven at 700 mA. A Ledil YASMEEN-50 secondary lens shapes the beam to a 60° ceiling wash, sized for a standard 2.4 m Australian residential ceiling.

Thermal management is entirely passive. The 6061 anodized-aluminium heatsink runs the full height of the rear face, sized through iterated benchtop testing. Custom LED mount printed in PAHT-CF (carbon-fibre-reinforced high-temperature nylon) for direct LED contact without softening. KSD9700 thermal switch as a hard 80 °C cutoff. Surface holds below 70 °C, comfortably under the cutoff.

Surface temp (max)< 70 °C

Thermal cutoff80 °C

LED life to L70> 50,000 h

Service life @ 8 h/day22 yr

Scatter plot of lamp surface temperature against ambient with a linear fit, R squared 0.871, predicting roughly 80 degrees Celsius at 40 degrees ambient — Thermal characterisation. Lamp vs ambient with linear fit (R² = 0.871) and mean ΔT = 48.1 °C ±1.8 °C over a 90 h run.

Stability testing

90+ hour burn-in. Power trend essentially zero (R² = 0.0001), comfortably above the L90 reference. The slight voltage drift is the constant-current driver compensating for the LED's V_f falling as the junction warms; both components on-spec.

Four-panel burn-in test chart over 90 hours: power degradation, voltage, current, and power residuals with L90 and L70 reference lines — Burn-in: 90 h continuous operation. Power trend ≈ 0, voltage stable, current within 1.305 to 1.310 A. L90 / L70 reference lines plotted.

Manufacture

The first batch was hand-assembled in Melbourne. Aluminium heatsinks are extruded; the PETG-HF outer housing is 3D-printed in-house for first-batch volumes. The product was designed for short-run assembly: every fastener is reachable with a single hex driver, every wire harness is keyed, and the bill of materials is short enough to fit on one page.

Bill of materials (excerpt)

№ItemReferenceQty

{[ ["01","LED module","Luminus CTM-14, 1800K, CRI 90+, 700 mA","1"], ["02","LED driver","MEAN WELL LDH-25-700, constant-current","1"], ["03","Secondary optics","Ledil YASMEEN-50, 60° beam","1"], ["04","Heatsink","6061 aluminium, anodized black","1"], ["05","Housing","PETG-HF thermoplastic","1"], ["06","LED mount","PAHT-CF (carbon-fibre-reinforced PA), 3D-printed","1"], ["07","Potentiometer","PTR90 100kΩ, integrated switch","1"], ["07","Thermal switch","KSD9700, 80 °C NC","1"], ["08","Power input","INJOINIC IP2721 USB-C PD trigger, 20 V","1"], ].map(([n, item, ref, qty]) => (

{n} {item}{ref} {ref} ×{qty}

))}

Stratco Cambodia Engineering and Design Department building, in-factory

ProblemRecruited by the Group CEO. Two engineers were working informally inside production, with no defined workflow, no documentation standard, and no live leadership visibility.

ApproachBuild the function as a documented operating model. Three-phase workflow with proportional effort weighting (Brief 10, Design 55, Sample/Prod 35), daily-refreshed SharePoint tracker as the single source of truth, embedded with production.

ToolsSharePoint and Power Automate for live tracking. IEEE-format Report Style Guide v1.0. Aluminium-printed Workflow Standard. 90-day project cadence.

Result21-project portfolio under oversight, 5 to production in 4 months. Leadership had real-time portfolio status without status meetings. Departmental infrastructure remains in place.

Brief

Stratco, a manufacturer of structural building products, had a limited engineering function in its Cambodia operation: design work happened informally within production, without a defined workflow, documentation standard, or live visibility to leadership. I was recruited by the Group CEO to formalise it: lead the team, define the workflow, and turn the project pipeline into something leadership could see.

Product imagery is withheld for commercial reasons. The artefacts shown below are anonymous engineering and QA detail. The substance of the role was the function-building.

By the numbers

21Projects in portfolio

05to prodLed to production in 4 months

02Engineers managed

06Stakeholders briefing work

Anonymised production output, from the 5 shipped

Finished pre-hung door, in factory aisle

Assembled fabrication detail, interior view

Departmental workflow

01. Brief

Project Definition Brief

Problem capture
Competitor / IP research
Acceptance criteria up front

02. Design

Design and engineering report

Detailed design
FEA / calculation
Report (Style Guide v1.0)

03. Sample / Production

First-article and handoff

DFM review
First-article QA
Production sign-off

The aluminium-printed workflow standard poster mounted in the department — Workflow standard, aluminium-printed, mounted in the department

Team and management

Two design engineers in the team, one junior. Defined the workload distribution and the design review cadence, trained the team on the new procedures and walked them through the workflow standard, and stepped into review on the higher-priority briefs. The team's output during the period included five products that reached production.

Operating cadence: worked between the dedicated D&E office and the factory floor. Fortnightly portfolio updates to Australian executives across five business units. Standard cadence of 90 days per small project.

Project management infrastructure: a daily-refreshed SharePoint tracker via Power Automate, the engineering report style guide, and the workflow standard shown above. Leadership saw current portfolio status without calling a meeting.

Engineering and Design Department office, title sign over the desk, planning whiteboard at the back

Featured technical work: Universal Steel Testing Rig

The one Stratco project shareable in full: the universal lateral-load testing rig. Cambodia previously relied on the Adelaide facility for testing HandiMate sheds, with weeks of turnaround per validation. The brief: commission an on-site rig for HM1 to HM10 against AS/NZS 1170.2 N1 to N4 wind requirements.

Black-steel SHS 75 x 75 x 4.5 mm primary frame, EA 30 x 30 x 3.5 mm bracing, M16 chemical-anchored to slab. 6.0 x 3.0 x 2.7 m envelope, 817.4 kg. Downgauged from 5.0 mm to 4.5 mm wall stock for local availability; released design is a 64% displacement improvement on the initial concept. Static FEA in Inventor 2026.2, peak field stress 35.6 MPa (8% of 300 MPa yield). Approved N1 to N4. Calibration protocol issued for the maximum-height load position.

Half-fabricated rig frame in the workshop

Von Mises stress visualisation at 10 kN proof load — Von Mises stress (MPa)

Displacement field at 10 kN proof load — Displacement (mm)

Safety factor distribution at 10 kN proof load — Safety factor

Assembly mass817.4 kg

Envelope6.0 x 3.0 x 2.7 m

Equipment rating10 t / 98.1 kN

N4 operating load6.86 kN

Peak stress at proof8% of 300 MPa

Displacement vs initial64% better

Wind region operating loads (AS/NZS 1170.2, long side)

Region Pressure Strength Serviceability

{[ ["N1", "0.66 kPa", "2.12 kN", "1.02 kN"], ["N2", "0.92 kPa", "2.95 kN", "1.03 kN"], ["N3", "1.44 kPa", "4.62 kN", "1.56 kN"], ["N4", "2.14 kPa", "6.86 kN", "2.31 kN"], ].map(([r, p, s, sv]) => (

{r} {p} {s} {sv}

))}

Approved N1 to N4 across all three load-application heights.

N4 displacement by load-application height

3.11mmLow height. Acceptable.

5.39mmMid height. Acceptable with note.

8.16mmMax height. Calibration protocol issued.

Seven Robotic Hand prototype variants in a line, showing four years of mechanical iteration

ProblemRobotic articulation is hard. Assistive navigation aids stop at the white cane's range. Two questions, one multi-year project: a tendon-driven hand mechanism from 2020, then a wearable haptic glove for blind navigation as the 2023 honours pivot.

ApproachOne prototype per engineering question. Seven hand prototypes across four years. The honours pivot took the hand-tracking and haptic findings forward into a wearable, validated through field interviews with white-cane users.

ToolsFusion 360 for design. Blender for visualisation. Arduino and Teensy stacks. In-house PCB design. FDM 3D printing for prototypes. Field interviews with white-cane users for design research.

ResultFirst Class Honours, RMIT (2023). Seven hand prototypes, one smart-glove submitted as the honours capstone. Engineering vocabulary built here: actuation, sensor selection, prototyping cadence, design research methods.

Origin and arc

Started in 2020 as a personal research project on tendon-driven articulation. Visible joints, exposed pulleys, fastener-rich. Seven prototypes across four years, each one closing one engineering question and opening the next.

In 2023 the work pivoted. Rather than the next hand prototype, the hand-tracking and haptic feedback findings evolved into a wearable smart-glove for blind navigation, submitted as my final honours project. First Class Honours, RMIT.

This is the foundational R&D piece in my engineering practice. The vocabulary built here: actuation kinematics, sensor selection, prototyping cadence.

Mechanism

Tendon-driven actuation with the actuator pack housed in the forearm. Each finger has its own tendon routed through low-friction guides and pulleys, with return springs in the dorsal side. Wrist rotation handled by a separate gear stage at the forearm interface.

Robotic Hand mechanical schematic with cyan tendons highlighted

Clean white render of the Robotic Hand, three-quarter view

CAD detail of the wrist actuator pack and gear linkage

Honours pivot: electro-tactile glove for blind navigation

Project Vision Sixth Sense (2023). A wearable glove for blind and low-vision users to navigate physical space without a white cane. Time-of-flight sensors on the dorsal surface and at the fingertips map nearby objects; vibration motors over the metacarpals translate that spatial map into haptic feedback under the skin.

Sensor stack: VL53L5CX multi-zone ToF, Grid-EYE infrared thermal array, OLED status. PCB designed in-house. Microcontroller (ATtiny vs Teensy) and battery sizing driven by wearable form-factor and haptic motor count.

Design research: field interviews with white-cane users on existing aids and acceptability (social acceptability, interference with object manipulation, design language reading as assistive technology versus medical device). Two glove design directions prototyped to render and physical sample.

Hand-drawn design exploration page for the smart glove with componentry list

Smart glove render, palm view, showing sensor mount and finger TOF positions

Smart glove render, back view, showing haptic motor mounts

Status

Hand work sits in long-running R&D. Most recent active iteration (2024) was on hand-tracking with twin BNO055 IMUs in the forearm and wrist for joint-angle inference without external cameras. The honours glove work is documented and not actively being prototyped further.