PhD in smart-building energy modelling | Physics-informed transfer learning

PhD in smart-building energy modelling | Physics-informed transfer learning

Quick Decision

This four-year, fully funded doctoral position, as stated in the captured vacancy published on 5 November 2026, focuses on transferring physically informed building-energy models to new buildings with little calibration data. It is a strong match for applicants with quantitative master's training and practical machine-learning, system-modelling or control experience who want to work with real building data.

Two points need early confirmation. The vacancy does not state salary or contract category, while the European Commission lists the SMARTWIN grant period as 1 January 2027–31 December 2029, shorter than the advertised four-year PhD. This does not mean the fourth year is unfunded; ask what budget and contract cover all 48 months. Also confirm the doctoral specialty, target buildings/data and division of work from the related safe/preference-aware MPC post. The two captured ads overlap substantially in their task lists, and the grant record does not allocate methods to individual students.

Research Problem and Doctoral Work

What Is Unsolved and Why It Matters

Building-energy models often need extensive building-specific data and can produce predictions that fit observations but violate physical expectations. The research challenge is to reuse what models learn from one or more buildings when adapting to another building with only a small calibration set, while representing thermal/HVAC processes plausibly and identifying unreliable predictions. This matters if forecasts are later used for energy decisions or control: a model that works on one building may fail when weather, occupancy, metadata or operating conditions change.

Recent reviews identify data scarcity, physically implausible outputs and weak generalization as recurring barriers in building-energy modeling. A 2025 review discusses how physical knowledge can enter through training losses and model architectures, but does not establish which approach will work best on TalTech data. The vacancy itself asks the doctoral researcher to study cross-building transfer, physical constraints, uncertainty and out-of-distribution conditions. Review of physics-informed ML for building energy modeling, Transfer Learning of Surrogate Models for Building Energy Prediction

Work, Methods, and Potential Outputs

The captured vacancy describes these candidate-level tasks:

• Develop building thermal and energy models that combine physical information with data-driven learning.
• Examine grey-box models, physics-informed neural networks, structured Gaussian processes and neural state-space models.
• Integrate energy-use and indoor-climate measurements with weather, occupancy and building metadata.
• Test transfer from models trained across multiple buildings to a new building with limited calibration data.
• Quantify uncertainty and assess reliability, including behavior under distribution shift.
• Validate against TalTech campus-building data and experimental facilities.
• Build reusable interfaces for later model-predictive-control and explainable-AI work, and produce open software, benchmark data and reproducible workflows; publish papers and present at conferences.

The posting names campus and experimental-facility data but does not identify building addresses, sensor channels, data volume, access approvals or a specific benchmark. TalTech describes relevant experimental capability, but facility presence does not establish a particular resource will be assigned to the candidate. One analyst-suggested design would compare zero-shot and few-shot transfer in a leave-one-building-out evaluation, with calibration budgets and uncertainty checks fixed in advance. That is a possible research angle, not an advertised deliverable.

SMARTWIN's official project description covers a broader capacity-building programme: physics-informed meta-learning, safe learning-supported model-predictive control, preference-based Bayesian optimization, researcher mobility and training, plus a planned smart-building control laboratory. This position foregrounds transfer learning; the related posting foregrounds safe and preference-aware MPC. Their task lists overlap substantially, however, and the public grant record does not assign distinct or exclusive work packages to either doctoral researcher. Do not assume one candidate owns all model development and the other all control work. SMARTWIN project fact sheet

School, Programme and Field Context

TalTech's School of Engineering offers the Engineering Sciences doctoral programme. The current curriculum record specifies 240 ECTS over eight semesters and a Doctor of Philosophy award under a selected specialty; entry requires a master's degree or recognized equivalent. Building and Civil Engineering and Architecture is plausible given the host group, but the vacancy does not name the enrolment specialty. TalTech's PhD admissions route distinguishes Early Stage Researcher employment from other doctoral forms. Confirm the route, specialty and degree registration for this position. Joint supervision with TU Darmstadt and possible collaboration with EPFL do not establish a joint or dual degree.

TalTech announced 600th place in the overall QS World University Rankings 2027. That is not a subject ranking for building-energy research. The more relevant setting is the School's established Nearly Zero Energy Buildings Research Group and its building-physics/HVAC work, plus a test facility with interior/exterior climate chambers and an assembly for studying building components and HVAC equipment. FinEst Centre for Smart Cities adds an interdisciplinary applied-research context. The vacancy names both the centre and the nZEB group as collaborators. Engineering Sciences doctoral curriculum, TalTech School of Engineering doctoral studies, TalTech QS 2027 announcement, Nearly Zero Energy Buildings Research Group, Zero Energy and Resource Efficient Smart Buildings centre, FinEst Centre for Smart Cities

Supervisor and Lab Assessment

Supervisor Research Directions

The captured vacancy names Avleen Kaur Malhi as principal supervisor and Tuule Mall Parts as co-supervisor. TalTech lists Malhi as an Adjunct Professor at FinEst Centre. Parts is linked to the nZEB group and has worked on building energy, control and modelling. The current TalTech contact card for Parts marks her senior-researcher employment contract as suspended; the posting still names her as co-supervisor. Ask the hiring contact to confirm her formal role and availability rather than drawing a conclusion from the directory note. Avleen Kaur Malhi — TalTech contacts, Tuule Mall Parts — TalTech contacts

Recent Important Publications

• Huotari, Främling and Malhi (2024), “Machine Learning Applications for Smart Building Energy Utilization: A Survey.” The survey organizes ML use across energy management/control, grid interaction, maintenance/security and personalization, and discusses open questions. It gives context for the principal supervisor's smart-building ML work; it is a review, not a TalTech transfer experiment. Publication
• Parts et al. (2023), “Wax actuator’s empirical model development and application to underfloor heating control with varying complexity of controller modelling detail.” The team calibrated actuator/valve behavior and compared controller models with increasingly detailed parameters in a simulated room. Deadband and PI settings influenced energy and temperature results more than actuator modeling. This illustrates why physical/control detail matters, but it does not study transfer learning. Publication
• Matson et al. (2025), “Building Potential Energy Savings Estimation Through Portfolio-Based Modeling.” The paper uses hourly consumption from 15 buildings to identify patterns, cluster buildings and estimate savings potential, with shopping malls as a case study. It shows adjacent group experience with multi-building energy data, not a physics-informed cross-building transfer model. TalTech Research Portal record

A 2025 IBPSA paper tests transfer between building-energy simulation surrogates and examines how much target data is needed; a separate 2026 paper examines transfer across climates and building types. These are methodological precedents, but using simulation surrogates differs from the vacancy's measured campus data. Earlier research on deep transfer for smart-building thermal dynamics is a direct comparator, though outside the recent-paper window. IBPSA transfer-learning paper, 2026 cross-building transfer study, Earlier smart-building thermal transfer study

Lab, Resources, and Collaboration Environment

TalTech's nZEB group works on building physics, HVAC, energy performance and renovation. The centre's published facility description includes interior and exterior climate chambers and a test assembly for studying building components and HVAC under controlled conditions. SMARTWIN is coordinated by TalTech with TU Darmstadt and EPFL as partners; the grant description includes mobility, joint training and a planned control lab. These are project-level arrangements and institutional capabilities. The vacancy mentions possible stays at TU Darmstadt, EPFL collaboration and campus/experimental data, but does not specify duration, budget, access or an individual schedule.

Funding, Application, and Living Cost

The captured vacancy states four years of full funding and an application deadline of 5 December 2026, 23:59 Europe/Tallinn. It gives no salary figure, contract terms, employment entity or start date. SMARTWIN's official grant period is 1 January 2027 to 31 December 2029. For a start in or after 2027, the four-year PhD extends beyond the grant's listed project period. The vacancy nevertheless says four years fully funded; request written confirmation of the funding route and contract coverage through month 48.

TalTech's current funding procedure guarantees a full-time Early Stage Researcher at least EUR 2,300 gross per month and says salary support can cover nominal doctoral study up to 48 months. This applies only if the candidate is hired in that category; the vacancy does not confirm ESR status, so treat the amount as conditional rather than the advertised salary. TalTech says doctoral study carries no tuition. Its general admission page sets a EUR 100 one-time qualification-review fee from 1 June 2026 for most applicants, with listed exemptions. TalTech ESR funding procedure, TalTech PhD admissions and fees, TalTech doctoral scholarships, SMARTWIN grant dates and scope

Tallinn housing estimates depend on room type. TalTech's current Akadeemia 5 listing gives EUR 296/month for a single room and EUR 348/month for a private dormitory flat, utilities included; it also lists waiting times of one month to a year. These are student residences, and eligibility for an employed doctoral researcher is unknown. July 2026 local reporting puts a TalTech-area double-room bed at roughly EUR 160–170 plus utilities, and nearby small flats around EUR 400–650; newer units may cost more. The national student information portal estimates non-housing costs at EUR 300–600 per month. These are planning ranges, not a guaranteed employee budget. TalTech Akadeemia 5 rates, 2026 Tallinn housing report, Study in Estonia cost estimate

Applicant Fit and Preparation

Required Background and Skills

The captured vacancy requires a relevant master's degree in electrical/control, civil/energy engineering, computer science, applied mathematics or a related field; grounding in ML, modelling, analytics or control; familiarity with time series, dynamical systems and/or building energy; programming and data-analysis ability (preferably Python, MATLAB or Julia); English at CEFR C1; analytical and mathematical problem-solving; independent and collaborative working; and readiness for visits, joint experiments, training and organisational tasks. It asks for a short essay on motivation and preferred research questions. The complete application checklist is not stated in the captured vacancy. TalTech's general admissions page uses B2 for foreign-applicant English documentation, but the vacancy's C1 requirement is stricter and should guide role fit.

Preferred Experience and Suggested Preparation

The posting lists physics-informed/scientific ML, hybrid models, transfer/meta/few-shot learning, PyTorch/TensorFlow/JAX, real sensor/weather/BMS data, system identification, building-energy modelling, HVAC/building physics, publications and open-source work as beneficial, not hard requirements. Useful preparation would be a reproducible time-series or dynamical-model project, a clear comparison of transfer with and without target-domain calibration, and a short discussion of uncertainty and physically implausible predictions. These are suggestions to demonstrate readiness, not added eligibility criteria. TalTech doctoral admission requirements, Engineering Sciences doctoral curriculum

Research Direction, Policy and Career Context

The EU's revised Energy Performance of Buildings Directive connects smart readiness with energy performance, indoor conditions, occupant and grid responsiveness, building automation, data, flexibility, privacy and security. Estonia's OSKA engineering forecast estimates nearly 4,700 new engineering workers will be needed from 2025 to 2030, including around 1,300 for employment growth and 3,400 for replacement demand; its forecast shortage depends on education and hiring patterns and is not a PhD placement rate. The European Construction Observatory reports that Estonia's draft building-renovation plan estimates EUR 8.6 billion in investment needs for 2026–2035. These sources show policy and engineering demand around building renovation, energy services, automation and data-driven performance assessment, but cannot predict a specific job or salary after the doctorate. EU Energy Performance of Buildings Directive, OSKA engineering outlook, European Construction Observatory: Estonia

Next Actions and Questions to Verify

Before Applying

• Request an employer-hosted vacancy link or written full terms, including required materials, interview stages, any fee exemptions and the expected start date.
• Ask which Engineering Sciences specialty and Doctor of Philosophy degree route will apply.
• Clarify whether the job is a full-time Early Stage Researcher appointment; request salary, contract entity, workload and employment duration.
• Ask which buildings, sensor/BMS channels, weather and occupancy data, and experimental facilities will be available, including privacy and data-use restrictions.

At Interview

• Ask how the transfer experiment is defined: source/target buildings, target-data budget, baselines, physical-consistency tests, uncertainty metrics and out-of-distribution checks.
• Ask how the work is divided from the related safe/preference-aware MPC vacancy. The captured descriptions overlap, and SMARTWIN's public record does not state candidate-level work packages.
• Confirm both supervisors' roles and availability, particularly the current directory status note for Tuule Mall Parts; ask what TU Darmstadt/EPFL participation means for the student.

Before Accepting an Offer

• Obtain written confirmation of funding and salary coverage for all 48 months, especially months 37–48 after SMARTWIN's scheduled end date.
• Confirm matriculation specialty, degree-awarding institution, co-supervision, mobility funding and any contractual work beyond research.
• Check whether doctoral employees may use TalTech housing and obtain current rates, deposits, utilities and availability.

Evidence and Sources

Vacancy-specific facts come from the full position notice, published on 5 November 2026. Public project, university, publication, policy, labour-market and housing sources used for the surrounding assessment are grouped below.

Project, university and doctoral training

• SMARTWIN — European Commission project fact sheet
• Admission to Doctoral Studies at TalTech
• Regulations of Tallinn University of Technology for Doctoral Studies
• Doctoral Studies — TalTech School of Engineering
• Engineering Sciences doctoral curriculum — TalTech ÕIS
• TalTech ESR funding procedure
• TalTech PhD scholarships and tuition information
• TalTech QS World University Rankings 2027 announcement

Supervisors, research groups and facilities

• Avleen Kaur Malhi — TalTech contacts
• Tuule Mall Parts — TalTech contacts
• Nearly Zero Energy Buildings Research Group
• Centre of Excellence for Zero Energy and Resource Efficient Smart Buildings
• FinEst Centre for Smart Cities
• Machine Learning Applications for Smart Building Energy Utilization: A Survey
• Wax actuator modelling and underfloor heating control study
• Building Potential Energy Savings Estimation Through Portfolio-Based Modeling

Field and market context

• Review of physics-informed ML for building energy modeling
• Transfer Learning of Surrogate Models for Building Energy Prediction
• Transfer learning across climates and building types
• Deep Transfer Learning for Thermal Dynamics Modeling in Smart Buildings
• EU Energy Performance of Buildings Directive
• OSKA engineering labour-market study
• European Construction Observatory country fact sheet: Estonia

Tallinn cost evidence

• TalTech Akadeemia 5 student-housing rates
• ERR report on Tallinn dormitory and rental prices, July 2026
• Study in Estonia cost-of-living estimates