available for consulting

I design the system. You build it.

I architect physical automation systems across the full stack: mechanical constraints, PCB requirements, firmware specifications, software integration. I specify how every layer connects and where each one can fail. The building is yours. The blueprint is mine.

system architecture mechanical design electronics firmware spec edge ML automation
pcb_test_arm.log
# architecture spec · PCB test robotic arm

## mechanical
arm 4 DOF, 200mm reach
eoat pogo pin array, 24 contacts
repeatability ±0.05mm

## electronics
controller Raspberry Pi CM4
drivers TMC2209 x4, UART mode
pogo_force contact 180g · spec needs 220g

## firmware spec
motion G-code over serial
probe_seq land · settle 80ms · measure
false_fail 3.2% on cold start

## data pipeline
storage SQLite · per board UUID
export CSV + REST endpoint

$ root_cause --issue false_fail
pogo force low on cold PCB.
fix: increase spring preload 40g.
layers specified
mech · elec · fw · data
engagement type
architecture review
first call
free · 30 min
availability
open
01
Specify First, Build Once

I define the full system before anyone builds it. Mechanical constraints, PCB requirements, firmware specifications, software interfaces and the tradeoffs between them. You get a blueprint that accounts for how every layer affects every other. Not a recommendations document. A specification.

02
Edge ML as a System Constraint

Inference on constrained hardware is not an ML problem. It is an architecture problem. I specify quantisation budgets, memory layout, inference window sizing and interrupt coexistence before a model is chosen. The model fits the system. The system does not stretch to fit the model.

03
The Interface Is Where It Breaks

Mechanical hands off to electronics. Electronics hands off to firmware. Firmware hands off to software. Each handoff is a place where unspoken assumptions live. I audit those interfaces, document the coupling requirements and make sure every discipline is building toward the same system, not their own version of it.

built from real frustration
mini-PLM. Because a mechanical revision should not silently invalidate your firmware.

Every integrated system I have worked on had the same failure mode: a design decision in one layer was made without knowing its downstream consequences in another. A BOM change that was never linked to the schematic it came from. A mechanical tolerance that silently invalidated a firmware timing assumption two months later. No tool tracked that coupling. So I built one. The demo is running a live build right now.

💿 1
2
DIY E READER
+
👁
Revision: v 3.0
Name Date Rev
DXF
BottomPlate_cnc_v1.0.dxf
Qty: 1₹300.00
3/9/2025 v 3.0
PDF
EPD_Controller_Flyer.pdf
Qty: 1
11/9/2025 v 2.0
STL
enclosure_body.stl
Qty: 1
11/9/2025 v 1.0
STL
display_bracket_lid.stl
Qty: 1
1/9/2025 v 2.0
PDF
pmic_schematic_v1.0.pdf
Qty: 1
4/9/2025 v 2.0
DXF
FrontPlate_cnc_v1.dxf
Qty: 1
3/9/2025 v 1.0
GX
epd_driver_comp.gx
Qty: 1
29/8/2025 v 1.0
STL
bezel_v2.stl
Qty: 2₹320.00
4/9/2025 v 1.0
MD
bom_rev3.md
Qty: 4₹899.00
29/8/2025 v 3.0

Your system has a fault path
you have not mapped yet.

30 minutes. I will tell you exactly where your architecture is fragile and what needs to change before it costs you a hardware spin. No pitch. No prep required. Just bring the system.

Book the free review free · 30 min · no commitment