whoami

$ whoami

juliusz mucharski

  • chemical engineer, bradford
  • BEng 2:1 · final year project 81%
  • polish citizen
  • UK settled status (right to work)
education

$ cat education.txt

MSc · lactic acid bioreactor

  • university of bradford · starting 2026
  • uncertainty optimisation of a kinetic / process model
  • same family as the LTS design: a real vessel, inputs that do not take their worst values independently

BEng (hons) chemical engineering

  • university of bradford · 2023 to 2026
  • 2:1 (overall 65%) · final year project 81%
  • process design · reactor design · heat & mass transfer
  • process safety · thermodynamics · numerical methods · uncertainty analysis
projects

$ cat projects.txt

LTS packed bed · ammonia plant · 81%

  • bradford · prof raj patel · 2025 to 2026
  • threw out a written 87% conversion the feed could not hit, designed to 85% under a ~86.5% ceiling
  • that choice is 15.5 t of catalyst instead of ~76 t, lab powder kinetics were not copied onto plant pellets
  • sized the vessel to ASME VIII, 1.80 m × 4.9 m, ΔP ~0.097 bar

event and regime models · live market data

  • python engine, Redis, a cockpit, models on live feeds, automated backtests
  • research then onto the live box if it holds, replay shadow and live share the same code

T2 laboratories runaway · IChemE SIESO

  • reconstructed the 2007 jacksonville explosion from the public CSB report
  • heat balance model and an article, uncertainty questions written before the run
  • a script recomputes the headline numbers from the committed data
skills

$ cat skills.txt

python · modelling

  • python, ODEs, parameter sweeps, Monte Carlo, live data through Redis
  • probability: Bayes, conditioning, when inputs are not independent
  • event and regime models on time series, automated backtests
  • backtest, shadow and live on the same code path
  • uncertainty: Monte Carlo, CVaR, chance constraints, Clopper Pearson bounds on spec violations
  • ask whether a number is real before trusting a point: intervals and certificates, not just a mean
  • global sensitivity (Sobol / Saltelli), bayesian calibration of scarce kinetics on a grid
  • process models: kinetics, heat balances, thermodynamic constraints
  • process safety and ASME VIII pressure vessel design
  • linux · git · english · polish
papers

$ cat papers.txt

in preparation · reactors, uncertainty, optimisation, control

  • tail risk optimisation of a batch reactive distillation: CVaR and chance constraints, then a certificate that the spec actually holds, the chance constrained pick can look fine until you replay it (winner's curse at the threshold)
  • which column you build can change once you score risk not just the mean, paired draws so the comparison is fair
  • certified closed loop: an absolute delivery rate mixes the controller with luck in the sample, pairing the same draws isolates the controller
  • criticality diagrams under uncertainty: class membership as a certificate not one temperature
  • whether a model is "conservative" depends on the event you certify not just the model
  • middle vessel operation under kinetic uncertainty, heat balances, reflux, real vessels, still writing, none of this is out yet