experience.

Where I have worked and what I shipped.

  • Building a neuro-symbolic platform for buyers in regulated domains who cannot act on an answer they are unable to trace back to a source.
  • Roughly an order of magnitude more usable structured output from identical input, by changing the method rather than the model.
  • Own the scope of the first release: what it has to do, what it deliberately will not do, and what has to be true before a feature is allowed to ship. Then build a large part of it.
  • Benchmarked candidate open models from 3B to 24B parameters on accuracy against cost per run, making model selection an explicit cost and performance decision rather than a default.
  • Settle the decisions that would force a rewrite if they are wrong before anything is built on top of them, and let the genuinely additive ones wait.
  • Put the evaluation harness outside the code it scores, on the argument that a harness sharing its subject's import graph has stopped being an independent measurement.
  • Walked participants disputing their energy, water and waste badges through the methodology behind each flag. Roughly 50% of them went back and amended their own reported values, which improved the dataset rather than just closing the ticket.
  • Co-designed an IQR-based anomaly detection module in Python and R, deployed inside the GRESB Real Estate product as a live in-product workflow, shipped to paying customers rather than kept as an internal script.
  • Wrote PRDs and feature specs, and partnered with Product on ESG Portal prioritisation.
  • Validated building performance data covering energy, water, waste and greenhouse gas against GHG Protocol-aligned rules, across 500+ entities.
  • First point of contact on the support queue alongside Customer Success, working assessment and technical-solution tickets in Front.
  • Built a sustainability assessment tool for early-stage startups, aligned to DIN SPEC 90051-1.
  • Advised more than ten founders on turning quantitative findings into product decisions, and on sustainable design and material choice.
  • Cut material waste by 20% by moving clients to PLA alternatives.
  • Earlier, as Student Assistant from Nov 2023 to Aug 2024: startup prototyping, CNC machining and 3D printing.
DIN SPEC 90051-1/
Life Cycle Assessment/
CAD/
3D Printing
  • Founding team member. Led the 3D printing and design team as Robotics Engineer from Jan to Jun 2021, after joining as Startup Associate on the incubation side in 2020.
  • Owned the loop from ideation through iteration to delivery of robotic kits for institutional clients.
  • Ran client communication and market research for those kits.
CAD/
3D Printing/
Robotics/
Market Research