Skip to content
Husain Bootwala

01/ 08

Mechatronics & Systems Engineer · Singapore

I design precision hardware and write the software that runs it.

Four years at KLA, taking semiconductor inspection equipment from NPI through production. Precision stages, motion control, DV/PV validation, and the field failure investigations that follow. Alongside that I build finished products on my own: a home-automation appliance with a local AI brain, a real-estate underwriting platform, and a spacecraft telecommand and telemetry stack.

4+ yrs
NPI & product engineering
50%
Test bench hardware cost reduction
$30k
Saved per field failure incident
50 hrs
Tool downtime prevented per incident

Readiness

07 channels

Where each project stands today, and what more time would buy.
  1. 1Concept designed, not running
  2. 2Prototype runs on the bench
  3. 3Pilot real users, limited
  4. 4In use relied on routinely
  5. 5Production deployed and supported
What the stages mean
  1. 1Concept designed, not running
  2. 2Prototype runs on the bench
  3. 3Pilot real users, limited
  4. 4In use relied on routinely
  5. 5Production deployed and supported
ChProjectStage todaySolves todayWith more time
01Mobile Particle Forensics RobotPrototype, stage 2 of 5. Motion stack runs in simulation onlyPlans one arm-plus-rail motion in sim and exports it as URScript a UR5e can run
  • Macro photo of a silicon wafer's grid of iridescent dies
    Finding particle sources
  • An industrial robot arm working inside a fenced production cell
    Safe reach near costly optics
  • A machine head with a red nozzle poised over a circuit board
    Repeatable inspections
02NexopPrototype, stage 2 of 5. Runs on a Jetson hub; beta not startedControls Zigbee, KNX and Home Assistant devices from one local hub, by chat or app
  • A row of DIN-rail circuit breakers in an open distribution board
    Pro KNX installs with no AI layer
  • A small round smart speaker on a wooden desk
    Hands-free control for older residents
  • A wall of electricity meters
    Energy waste nobody notices
03Spacecraft TC/TM LinkPrototype, stage 2 of 5. Complete bench demo; no flight hardwareSimulates a satellite power unit and checks every telemetry reply against safe limits
  • Software waiting on flight hardware
  • An oscilloscope and signal generator on a lab bench
    Fault paths unsafe to trigger for real
  • Rehearsing fault response
04HexstateIn use, stage 4 of 5. live · used daily on its own repoRuns a team of AI agents on your projects while you do the real-world work
  • Solo founders doing every role alone
  • Small teams with no ticket discipline
  • AI code changes that nobody reviews
05Laundry TrackerIn use, stage 4 of 5. In use at a laundry in AfricaTags every garment so the laundry using it knows what is in, out, and overdue
  • A row of front-loading washing machines in a laundromat
    Garments lost between branches
  • Customers ringing to ask if it's ready
  • A torn white T-shirt held together with safety pins
    Damage claims with no photo
06NextDealIQPrototype, stage 2 of 5. Works end to end; waitlist openRuns a deal as rental, BRRRR, flip or CMA and builds the client report from it
  • Follow-ups that slip after a showing
  • Viewings booked by phone tag
  • Teams with no shared deal pipeline
07Pharmacovigilance Productivity AIConcept, stage 1 of 5. Research done; no code, no direction yetMaps a drug-safety team's day and scores 72 places AI could help, step by step
  • Tall stacks of paper case files with yellow tabs in an office
    Re-keying every case field
  • Medical textbooks with bookmarks beside a stethoscope
    Literature screened by hand
  • Cases bouncing back from QC

On how I build

Writing code stopped being the difficult part some time ago. What separates engineers now is how well they run AI: the rules an agent has to work inside, whether the documentation still describes the code a month later, and which approvals are reserved for a person. Each project below documents how it was built as well as what it does.

Selected work

07 channels
2026 —

Mobile Particle Forensics Robot

A cobot that reaches inside semiconductor equipment to find where contamination starts

Particle contamination inside a wafer inspector is invisible until it kills yield, and finding its source means a human with a flashlight and a hunch. This is a UR5e-class arm on a rail base carrying a UV inspection head, with an LLM agent that proposes inspection sequences and executes them behind human confirm-gates.

  • UR5e / URScript
  • ROS 2 Jazzy
  • MoveIt 2
  • URSim (PolyScope)
  • Gazebo
  • +6
Active R&DHardware
2025 —

Nexop

A smart-home appliance that learns a household's routine and proposes the automations for it

Smart homes break down at the last step: the devices work, but writing the rules is a chore nobody keeps up with. Nexop learns recurring patterns from local device events and proposes automations the homeowner can approve, edit or ignore. Native Zigbee and KNX, Home Assistant for everything else, all of it running on a Jetson in the house rather than a vendor's cloud.

  • FastAPI
  • Next.js
  • TypeScript
  • PostgreSQL
  • Redis
  • +8
Active R&DHardware + Software
Apr 2026

Spacecraft TC/TM Link

A two-sided EGSE bench for a satellite power distribution unit

Built under contract, through my company Habzeno, for a satellite company that needed a bench to test its parts. Ground support equipment for spacecraft integration: a subsystem simulator so ground software can be exercised before flight hardware exists, and a ground client so the subsystem can be exercised before the ground station does. The client commands a load, decodes the returned telemetry, evaluates it against safe operating limits, and logs every decision, including the fault case that you cannot safely produce on real hardware.

  • Python
  • Binary protocol design
  • Sockets
  • Docker Compose
  • CSV telemetry logging
ShippedSoftware
2026 —

Hexstate

A whole AI engineering team that runs on the Claude Code or Codex plan you already pay for

A desktop app that gives you a team of AI agents: a manager, a product manager, engineers, a reviewer and a personal assistant. They talk to each other in shared channels, and you talk to them like employees. Every change goes through a ticket and waits for your sign-off. You do the real-world work; Hexstate does the building, even while you sleep.

  • TypeScript
  • Claude Code CLI
  • Codex CLI
  • Electron
  • React 19
  • +8
LiveSoftware
2026

Laundry Tracker

QR-tagged garment intake for a laundry that was running on memory and paper slips

A working laundry loses garments, and every lost garment is an argument with no evidence on either side. This tags each item with a printed QR code at intake, times it through the shop by household, prints a receipt and drafts the customer's WhatsApp message. It ships as a desktop app because the shop counter has one PC and no IT support, and it is in use at a laundry in Africa.

  • FastAPI
  • Python
  • SQLAlchemy
  • SQLite (desktop) / Supabase Postgres (hosted)
  • Jinja2
  • +4
ShippedSoftware
2026 —

Pharmacovigilance Productivity AI

Mapping a regulated industry's working day before committing to a product

Drug safety teams process adverse event reports under legal deadlines, and the obvious AI answer (automate the case assessment) is the one regulators are least willing to accept. So I mapped the case lifecycle and ten roles first, then pinned and scored every place AI could add value against those maps. I am not ready to choose a direction yet.

  • Domain research
  • Process modelling
  • Regulatory analysis
  • Opportunity scoring
  • Agentic research pipelines
In discoverySoftware