TU Delft PhD: Bayesian Modeling and Counterfactual Methods for Nanoparticle Transport in Human Tissue

TU Delft PhD: Bayesian Modeling and Counterfactual Methods for Nanoparticle Transport in Human Tissue

Recruitment and policy information reflects the October 2026 material; current availability and individual terms require confirmation. The research discussion retains the stated earlier selected-paper reading limits.

Decision in brief

This computational doctorate has unusually explicit requirements: a quantitative MSc, ODE/dynamical-system modelling, Bayesian inference in a probabilistic programming framework, and strong Python with a code repository. It sits in TU Delft’s Faculty of Technology, Policy and Management (TPM) with Dr Sepinoud Azimi, within SYNERGIE, Horizon Europe grant 101288129. Biology, chemistry, toxicology or nanomaterials experience is expressly not required. TU Delft official vacancy 3924: nanoparticle transport

The employer gives 29 November 2026 in both the application paragraph and submission field, whereas the listing header states 30 November; applicants should follow the employer application date and confirm the portal. The exact closing time is unverified. The intended start is January 2027. Funding is employment: 1.5 years initially, a go/no-go assessment within 15 months, then 2.5 years subject to satisfactory performance. First-to-fourth-year pay is €3,204–4,051 gross monthly, based on 38 hours. TU Delft official vacancy 3924: nanoparticle transport

The opportunity combines mechanistic inverse modelling with experimentally informed intervention design. The central risk is whether enough informative chip data arrive in time to identify the changing particles’ behaviour and validate proposed changes. A model’s recommended redesign must not be mistaken for demonstrated human safety or therapeutic efficacy.

Research problem and actual doctoral work

A nanoparticle can acquire a protein corona, aggregate and, depending on the material, partly degrade as it moves. Its initial size/composition may therefore fail to describe what reaches the next tissue. SYNERGIE aims to connect nine human tissue-on-chip models across uptake, circulation and storage; TU Delft leads the computational work package. The official job describes thirteen partners, including Fraunhofer, Roche, Karolinska, University Hospital Aachen, Mimetas, Ruhr Bochum and Maastricht. The exact nine-tissue list and current experimental readiness are not supplied. TU Delft official vacancy 3924: nanoparticle transport

The advertised tasks are to build a mechanistic model of inter-tissue transport, calibrate it using Bayesian inference, identify influential physical/chemical properties, and ask what smallest feasible change could improve undesirable behaviour and when that recommendation cannot be trusted. Outputs include open-source software and support for regulatory dialogue. Departmental teaching and BSc supervision are also duties. A postdoc, related PhDs, experimental partners and DelftBlue computing are named resources; CPU/GPU allocation and data-delivery schedules remain to be agreed. TU Delft official vacancy 3924: nanoparticle transport

The statistical difficulty is separating changing particle states from the signals observed in chip data. Several parameter combinations or model structures may explain the same observations; a narrow posterior does not establish that the model is correct. Identifiability, measurement uncertainty and the scope of independent evidence therefore determine how much can be concluded. These are scientific questions for the doctorate, not evidence that the consortium has already resolved them.

Counterfactual conclusions require a stronger interpretation than accurate prediction. A conditional association or feature attribution does not establish the effect of a hypothetical change, especially outside the observed material range. Agreement between model structures and the limits of available observations should bound any claim. No particular delivery improvement or human safety conclusion follows from a fitted transport model. The supervisor’s recent work makes these distinctions particularly relevant. Kibria et al. 2023: predictive nanoparticle simulation surrogate, Jansen-Storbacka et al. 2026: evolutionary therapy models, Azimi 2025 review: intelligent nanoparticle design

University, subject and doctoral training

TU Delft employs and awards the doctorate. TPM’s engineering, modelling and policy environment is a specific fit for software intended for scientific and regulatory use; it is not a clinical medical degree. The consortium’s institutions supply research collaboration, without an advertised joint degree. The current profile lists Azimi as an associate professor of Explainable AI for Medicine. TU Delft official vacancy 3924: nanoparticle transport, Sepinoud Azimi current TU Delft profile, TPM Graduate School

THE reports 50 overall in 2027, 16 in Engineering in 2026, and 45 in Computer Science in 2026. Those categories describe institutional and broad subject strength, not nanotoxicology, this laboratory or supervision outcomes. THE Delft 2027 overall and 2026 subjects

The current admission page links the 2023 Doctoral Regulations and implementation decree edited in 2024. Its educational component requires 45 GS credits: 15 each in transferable, research-related and discipline-related skills, including at least 5 from learning on the job. One GS credit is 12 hours, so it must not be relabelled 45 ECTS. TPM and the university Graduate School provide training and support; specific courses, mentor, formal promotor and supervision cadence should be recorded for this appointment. The contractual 15-month decision needs a practical milestone plan, not only a calendar date. TPM Graduate School, Doctoral regulations and implementation decree 2023/2024

Supervisor’s recent papers and what they demonstrate

Azimi’s profile connects formal computational modelling, explainable/generative AI, nanoparticle design and biological barriers. NAP4DIVE work and a 2026 consortium workshop show a relevant European network, but NAP4DIVE and SYNERGIE are different projects. Existing collaborators are not a guarantee of placements, chip access or particular publication outputs. Sepinoud Azimi current TU Delft profile, NAP4DIVE consortium 2026 workshop record

• Kibria et al., Scientific Reports 2023, 13:547. A hybrid time-series/surrogate approach predicts solvent-accessible surface area (SASA) from partial molecular-dynamics trajectories. The selected simulation study covers 107 designs; limited diversity and information loss in structural descriptors constrain interpretation. Feature explanations are not direct evidence about feasible physical changes. SASA is a computational proxy, not measured cancer-treatment efficacy. Azimi jointly supervised with Lafond and shared concept/method planning and writing; Kibria wrote the main pipeline/training scripts. The earlier analysis used selected publisher-PDF methods, discussion, limitations and contributions. Kibria et al. 2023: predictive nanoparticle simulation surrogate

• Jansén-Storbacka et al., Scientific Reports 2026, 16:7442. The study compares 26 two-population differential-equation models of erlotinib-treated lung-cancer dynamics. After selection, 13 patients have sufficient longitudinal tumour observations; volumes use a spherical approximation. Different structures can fit observations yet imply different intervention outcomes. Modelled time-to-progression improvements are simulation results, not benefits observed in an adaptive-therapy trial. Azimi’s stated roles include methodology, formal analysis, drafting and supervision; software is credited to other authors. The version underlying the earlier analysis is the publisher’s accepted, unedited manuscript, not the final edited PDF. Differential-evolution fitting here should not be described as proof of Bayesian inference. Jansen-Storbacka et al. 2026: evolutionary therapy models

• Grolleman et al., Scientific Reports 2023, 13:18374. Cell populations in different biological and environmental contexts show time-dependent morphology and mechanics, including observations at 24 versus 48 hours. It demonstrates context-sensitive biological data analysis. Azimi and Raza implemented analytic tools; others designed/performed the experiments and supervised. This is not nanoparticle transport, nor evidence that she led the wet laboratory. The earlier analysis used selected final institutional-PDF methods/results, discussion and contribution sections. Grolleman et al. 2023: cell mechanical homeostasis

The 2025 sole-author nanoparticle-AI review supports her synthesis of formulation, barriers and interpretability challenges, but its available reading scope is limited to the institutional abstract and metadata. It is not an additional primary validation study. No verified lab completion rate, placement census or daily mentoring schedule was found. Azimi 2025 review: intelligent nanoparticle design

Pay, fees and the Delft budget

The official July 2026 P0–P3 salary table gives €3,204/3,706/3,873/4,051 gross monthly; the vacancy independently confirms the first/fourth amounts plus 8% holiday allowance and 8.3% year-end bonus. These are not equal monthly net cash. The agreement currently covers July 2026–June 2027, and later negotiated rates are unknown. Ask HR for personal payroll/pension/tax calculations rather than assuming a tax-relief ruling. The advertised header also shows 36–40 hours, while pay explicitly uses a 38-hour full-time contract; settle the chosen schedule in the contract. TU Delft official vacancy 3924: nanoparticle transport, UNL salary scales effective 1 July 2026, UNL agreement Dutch Universities 2026–27

The university’s public fee page lists €11,000 for the entire PhD programme, plus a variable annual bench fee. It says fees/payer should be agreed before the start, bench fees are charged to the institute where funded, and faculty waivers may apply. This does not establish that this employee owes €11,000 personally, or that every research expense is automatically waived. Confirm the project’s tuition/bench/training/publication/computing payer; do not add international taught-master’s tuition. TU Delft doctoral fees and funding

The official international-staff housing table, July 2026, quotes €785–980/month for a single studio and €800–1,100 for a couple, including utilities, internet and municipal taxes. This is initial accommodation for up to six months, subject to availability, with a €285 service fee and specific refund conditions. Children and pets are excluded from those DUWO units. It is neither guaranteed housing nor a long-term market quotation. TU Delft international PhD short-stay housing

For an initial single-person scenario, €1,500–2,000/month (€18,000–24,000/year) can be used as an analyst planning range, combining the dated housing indication with a contingency for food, medical insurance, transport and daily costs. It is not an official university total. The staff cost page’s worked example is explicitly September 2023, a starting postdoc with a non-earning partner and two children; its net salary and family rent must not be copied as this PhD’s 2026 budget. Obtain current insurance quotes and a lease, and allow for deposit, relocation and later housing. Dutch employee basic-health-insurance obligations remain relevant. TU Delft staff cost-of-living guide: dated example, Dutch government health insurance for employees, TU Delft international employee support

Requirements, Chinese applicants and application materials

Required: relevant quantitative MSc; demonstrated ODE/dynamical modelling; Bayesian inference with PyMC, Stan, NumPyro or a comparable framework; strong scientific Python and a repository. Preferred: optimization/control, causal/counterfactual reasoning, surrogates/Gaussian processes, PBPK/systems pharmacology, and packaging/documentation. Biology/toxicology background is not required, but willingness to learn experimental meaning matters. A reproducible small ODE/Bayesian study showing solver checks, parameter recovery and a held-out prediction is an analyst portfolio suggestion, not another formal requirement. TU Delft official vacancy 3924: nanoparticle transport

The university requires a Dutch master’s or foreign equivalent. IELTS Academic normally needs 7.0 overall/6.5 each; TOEFL tests from 21 January 2026 use 5 overall/4.5 each, while earlier tests need 100/22. Cambridge C1/C2 at 185 is an alternative. Results must be within two years and are requested later in selection, not necessarily with the initial application. Native speakers and, in principle, English-taught master’s graduates may be exempt, but the latter may still face a placement test. Chinese degrees can enter equivalence assessment; individual recognition and an English-taught-degree exemption are not automatic. TU Delft current doctoral admission requirements

Apply online by 29 November 2026, with CV, brief cover letter, one-page research statement and a link to code you built. Explain the intended work, fit, dynamical modelling and Bayesian experience. Two referees’ details with consent are requested if selected for interview, rather than two reference letters at initial submission. Email/post applications are excluded. Contact: s.azimirashti@tudelft.nl. TU Delft official vacancy 3924: nanoparticle transport

The job lists no nationality restriction or MSCA mobility rule. Horizon Europe funding alone does not create the 12-in-36-month MSCA condition. The advertised late-stage knowledge-security assessment examines supplied CV/letter and relevant affiliations/research sensitivity; it must not be turned into a blanket exclusion of Chinese applicants. Permit support and accompanying-partner services help relocation without guaranteeing immigration approval or partner employment. TU Delft knowledge-security risk assessment, TU Delft international employee support

Field outlook and staged verification

The European Commission adopted a non-animal chemical-safety roadmap on 1 June 2026, with recommendations spanning several regulated domains. This makes experimental/computational validation and regulatory communication relevant capabilities. It does not certify SYNERGIE’s model, remove all existing tests, or establish that an in-vitro transport prediction is a human exposure limit. EuropeanCommission 2026 roadmap towards non-animal safety

Potential routes include scientific software, Bayesian modelling, PBPK, nanomedicine data science and non-animal/regulatory methods. Clinical practice or product approval needs different qualifications/evidence. Dutch ICT data show both vacancies and a cooling market; they cannot provide this lab’s placement probability or a guaranteed industry salary. UWV Dutch ICT labour-market overview

Before applying: confirm the portal’s exact closing time, the individual qualification/English route and flexibility around the January start.

At interview: establish the actual nine tissues/material classes and first data delivery; model identifiability and independent validation; the assumptions and evidence limits of counterfactual conclusions; daily mentor/formal promotor; computing allocation, teaching and 15-month assessment milestones.

Before accepting: obtain written continuation criteria, fees/research-cost responsibility, personal net-pay calculation, permit timing and workable initial plus long-term housing. The most useful outcome is an agreed feasible first-year model and validation plan, not a promise that nine chips already reproduce a whole human body.

Evidence and sources

Project, institution and admission

• TU Delft official vacancy 3924: nanoparticle transport

• Sepinoud Azimi current TU Delft profile

• TPM Graduate School

• TU Delft current doctoral admission requirements

• Doctoral regulations and implementation decree 2023/2024

• THE Delft 2027 overall and 2026 subjects

• TU Delft knowledge-security risk assessment

• TU Delft international employee support

Recent research and field context

• Kibria et al. 2023: predictive nanoparticle simulation surrogate

• Jansen-Storbacka et al. 2026: evolutionary therapy models

• Grolleman et al. 2023: cell mechanical homeostasis

• Azimi 2025 review: intelligent nanoparticle design

• NAP4DIVE consortium 2026 workshop record

• EuropeanCommission 2026 roadmap towards non-animal safety

• UWV Dutch ICT labour-market overview

Pay, fees and living costs

• TU Delft doctoral fees and funding

• UNL salary scales effective 1 July 2026

• UNL agreement Dutch Universities 2026–27

• TU Delft international PhD short-stay housing

• TU Delft staff cost-of-living guide: dated example

• Dutch government health insurance for employees