Dr Tianheng Zhao. Delivered with ai transformation, Grounded in science, Built for people.

Hi, I'm Tian.

Scientist, deep-tech consultant, startup R&D lead, and now Lead AI Consultant. Different titles, same job: taking something that works in a lab or a demo, and making it work for the people who use it and pay for it.

Get in touchRead the story
---

01 · WHAT I BRING

Science, AI, and the people in between.

02 · STORY

Three careers, one job.

ACT I · 2012 – 2020 · SCIENCE

Colour without dye.

FIG. 01 — MORPHO BUTTERFLY WING · STRUCTURAL COLOUR

ACT I · 2012 – 2020 · SCIENCE

Colour without dye.

Did you know that almost every blue on a butterfly or a bird comes from nanostructure, not pigment? Plants do it too: in the Pollia berry, cellulose arranges itself into a repeating structure that reflects light as colour. My PhD was about borrowing that trick to make bright, metallic colours from plant ingredients. Alongside the PhD I co-founded WaterScope, a startup using computer vision to test drinking water, which was my start in AI. The lesson that stuck: brilliance comes from arrangement, not ingredients.

2012 – 2015
Monash University · BEng Materials, First Class
2015 – 2020
University of Cambridge · MPhil & PhD, photonic materials
2015 – 2023
WaterScope · co-founder, edge-AI water testing

ACT II · 2020 – 2025 · DEEP-TECH COMMERCIALISATION

Science people will pay for.

FIG. 02 — LAB → PILOT → PLANT

ACT II · 2020 – 2025 · DEEP-TECH COMMERCIALISATION

Science people will pay for.

A great result in the lab is a long way from a product someone will buy. Closing that gap is mostly unglamorous work: cost, yield, safety rules, and making sure it works on the hundredth batch, not just the first. I spent five years there. At TTP, a Cambridge engineering consultancy, I led client projects in medical devices, filtration and advanced materials, from the vision system for a medicine-dispensing robot to two granted patents. Then at Sparxell I went back to plant-based colour and scaled it from grams a day to kilograms, at a tenth of the material cost, for brands like LVMH, Moncler and Gucci. Most of the job turned out to be translation: scientists, engineers, investors and customers all need the same truth told differently.

2020 – 2023
TTP · deep-tech & AI consultant
2023 – 2025
Sparxell · Head of R&D

ACT III · 2025 – NOW · AI TRANSFORMATION

AI people actually use.

FIG. 03 — HUMAN IN THE LOOP

ACT III · 2025 – NOW · AI TRANSFORMATION

AI people actually use.

Most AI projects stall at the proof of concept. The model usually works; what's missing is the bigger picture: what impact it should have, and how it fits the people who'll actually use it. At SyndicateRoom, investor updates went out as the same email to everyone, and fewer than one in five were opened. I rebuilt them so AI turns thousands of company updates a month into a short summary of what each investor actually holds, with the financial regulations built in from the start. Opens went to around 70%. At Softwire, an accounting startup wanted AI at the heart of its product, with no clearer brief than that. Most of my job was getting everyone to agree what done looked like; seven weeks later it was live. On the side I run OutRung, an AI-assisted job search portal, so I still build, test and run these systems myself.

2025 – 2026
SyndicateRoom · Head of Data & AI
2026 – now
Softwire · Lead AI Consultant
2026 – now
OutRung · founder, AI-assisted job search

03 · CONTACT

Working on something similar?

If you're somewhere between a lab result and a product people use, in science, scale-up or AI, I'd like to hear about it. I'm also happy to give talks, compare notes, or coach you through a career change like mine.