9 CS PhD & Postdoc Openings from Active ML Labs — Week of May 10–17, 2026

Nine labs recruiting right now across ML theory, computer vision, NLP, geometric deep learning, materials ML, and responsible AI. All but one have published at NeurIPS, ICML, ICLR, CVPR, or ACL within the past two years — the one exception is flagged. Sorted by deadline, closest first.

Act now: deadlines by May 31

Responsible AI PhD positions — University of Vienna

PI: Prof. Martin Pawelczyk, Assistant Professor for Responsible AI, Faculty of Computer Science, University of Vienna, Austria 1

Position type: Multiple fully funded PhD positions (started recruiting Feb 2026)

Research area: Three interlocking tracks — (1) data-centric AI: machine unlearning, data attribution, privacy for LLMs; (2) AI safety: robust alignment in multi-agent LLM systems; (3) mechanistic interpretability of large models 2

Deadline: May 20, 2026 (3 days from today — apply immediately)

Location: Währinger Str. 29, Vienna, Austria

Recent top-venue publications: "In-Context Unlearning: Language Models as Few Shot Unlearners" (ICML 2024), "Weak-to-Strong Trustworthiness in Language Models" (ICML 2025) 3, "Efficiently Verifiable Proofs of Data Attribution" (NeurIPS 2025), "Train Once, Answer All" (ICLR 2026), "Machine Unlearning Fails to Remove Data Poisoning Attacks" (ICLR 2025) 1

Lab culture: The group was established in February 2026, making this a genuine ground-floor opportunity. Pawelczyk previously held a postdoc at Harvard and a PhD from the University of Tübingen. He also did a research stint at JP Morgan AI Research. He is an Area Chair for NeurIPS 2026 and KDD 2026. On the lab's philosophy: "Joining our team as an early member will allow you to shape our research culture and tackle high-impact projects from the ground up in a world-class research environment." 1 The group explicitly welcomes applicants from CS, mathematics, statistics, or ML backgrounds — coding fluency is required but can be developed during the PhD.

Apply: 2

PhD in explainable foundation models — INRIA Paris ⚠️

PI: Dr. Caterina Urban, ANTIQUE team, Inria Paris & ENS | PSL, PR[AI]RIE-PSAI consortium

Position type: PhD (3-year fixed-term contract), start date November 1, 2026

Research area: Formal specifications for foundation models — the project aims to extract logic-based "neural specifications" from internal attention-head representations of LLMs that are both interpretable and machine-verifiable 4

Deadline: May 25, 2026

Location: Paris, France (Inria Paris campus)

Top-venue publication status: ⚠️ Borderline — verify before applying. Dr. Urban's primary publication venue is formal methods (POPL — Symposium on Principles of Programming Languages — 2026: "A Logic for the Imprecision of Abstract Interpretations"). She organized a NeurIPS 2025 workshop (UrbanAI) and the job posting cites ICML 2023 background work, but a main-track NeurIPS/ICML/ICLR/CVPR/ACL publication by Dr. Urban for 2024–2026 was not confirmed during research. 5 Applicants whose primary criterion is "publishing at ML top conferences" should verify directly with the PI.

Lab culture: International environment with planned research visits to the University of Toronto (collaborating with Xujie Si, a leading researcher in specification learning). Inria benefits include subsidized meals, 7 weeks annual leave plus 10 RTT days, hybrid working, and flexible hours.

Apply: 4

PhD in mechanistic interpretability for LLM security — University of Copenhagen

PIs: Prof. Isabelle Augenstein (Full Professor, Deputy Head of Department for Research) & Prof. Pepa Atanasova (Tenure-Track Assistant Professor), Department of Computer Science, NLP Section, University of Copenhagen, Denmark 6

Position type: Fully funded PhD fellowship, 3 years, start September 2026 or earlier

Research area: Mechanistic interpretability methods to identify and counteract false-information attacks on LLMs; sits at the intersection of NLP, LLM security, and explainable AI 7

Deadline: May 31, 2026

Location: Copenhagen, Denmark

Recent top-venue publications: Augenstein: 3 papers at ACL 2025, 6 papers at ACL 2026; Atanasova: 3 papers at ACL 2025, EMNLP 2025 8 9

Lab culture: CopeNLU (Copenhagen Natural Language Understanding) has 20+ members and is part of Denmark's Pioneer Centre for AI, funded by the EU. Augenstein is an ERC Starting Grant holder, recipient of the Karen Spärck Jones Award for contributions to natural language processing, and a member of the Royal Danish Academy of Sciences and Letters. 6 Lab alumni have gone on to NVIDIA, ETH Zurich, Hugging Face, and Aalborg University. The group is explicitly international and has visiting researchers year-round.

The posting states: "I am recruiting a PhD student and postdoc for our new project 'A Mechanistic Framework for Mitigating the Susceptibility of LLMs to Learning False Information' funded by Independent Research Foundation Denmark." 6

Apply: 7

Postdoc in LLM factuality detection — University of Copenhagen

PIs: Prof. Isabelle Augenstein & Prof. Pepa Atanasova (same lab as above — CopeNLU) 8

Position type: Postdoc, 3 years, start September 2026 or earlier

Research area: Characterizing the spectrum of false information that LLMs produce at different lifecycle stages; developing mitigation methods; preventing false-information attacks 10

Deadline: May 31, 2026

Location: Copenhagen, Denmark

Notable: The posting mentions a collaboration opportunity with NVIDIA as part of an existing academic relationship — postdocs may apply to participate as an academic collaborator. 10

Lab culture: See Copenhagen PhD entry above — same lab, same PIs.

Apply: 10

Longer window: June and beyond

PhD in machine learning — TU Dresden

PI: Prof. Dr. Bjoern Andres, Chair of Machine Learning for Computer Vision, Institute of Artificial Intelligence, TU Dresden, Germany

Position type: Fully funded Research Associate / PhD student (4 years, salary group E13 TV-L — the standard German public-sector pay scale for research staff, roughly €55,000–€70,000/year pre-tax depending on experience)

Research area — choose one of three tracks: 11

ML + Image Analysis
ML + Combinatorial Optimization
ML + Automated Theorem Proving

Deadline: June 1, 2026

Location: Dresden, Germany

Recent top-venue publications: "Box Facets and Cut Facets of Lifted Multicut Polytopes" (ICML 2024), "Partial Optimality in the Linear Ordering Problem" (ICML 2024) 12, "A Sub-Problem Quantum Alternating Operator Ansatz for Correlation Clustering" (ICML 2025) 11

Lab culture: Small group embedded in TU Dresden's Institute of Artificial Intelligence, part of the DRESDEN-concept research alliance. The lab's publication record spans ICML, AISTATS, GCPR, TPAMI (IEEE Transactions on Pattern Analysis and Machine Intelligence), and Discrete Applied Mathematics — a clear signal of the group's mathematical rigor. The PI supervised PhD students with first-author papers at ICML and NeurIPS venues. Prior to TU Dresden, Andres was affiliated with the Max Planck Institute for Informatics, Saarbrücken. The position description emphasizes curiosity-driven basic research — candidates choose their own track within ML.

Apply: 11

Multiple PhD positions in computer vision — University of Luxembourg

PI: Prof. Djamila Aouada, Associate Professor, Deputy Head of SnT, Head of CVI2 Research Group, Interdisciplinary Centre for Security, Reliability and Trust (SnT), University of Luxembourg

Position type: Doctoral Researcher (PhD), 36-month contract, €41,976/year gross, multiple positions available 13

Research area: Computer vision across a broad portfolio — 3D understanding, visual generative modeling, multimodal learning (vision-language models), media forensics, anomaly detection

Deadline: Rolling (applications processed upon reception)

Location: Kirchberg Campus, Luxembourg City, Luxembourg

Recent top-venue publications: "Fine-Grained Reconstruction of Constrained CAD Models" (NeurIPS 2025) 14, "SpelsNet: Surface Primitive Elements Segmentation" (NeurIPS 2024), "CAD-SIGNet: CAD Language Inference from Point Clouds" (CVPR 2024) 15

Lab culture: CVI2 has 23 staff, spans a Computer Vision Lab and an unusual Zero-G Lab — a facility for simulating space-like proximity operations, used in satellite servicing and autonomous robotics research. Application domains include space systems, Industry 4.0, surveillance, cybersecurity, healthcare, and automotive. The lab page describes the group as "highly motivated, active members" with "advanced GPU computing resources." Strong computational infrastructure at SnT. 15

Apply: 13

Postdoc in geometric deep learning — Télécom Paris

PI: Prof. Jhony H. Giraldo, Assistant Professor, Télécom Paris (Institut Polytechnique de Paris), LTCI — Information Processing and Communications Laboratory, Multimedia team; ELLIS Society member 16

Position type: Postdoc, starting spring/summer 2026

Research area: Geometric deep learning — graph neural networks and higher-order (simplicial / cellular) networks — with applications in computer vision and graph signal processing

Deadline: Rolling (posted May 2026)

Location: 19 place Marguerite Perey, Palaiseau, France (near Paris)

Recent top-venue publications: "Continuous Product Graph Neural Networks" (NeurIPS 2024) 17, "Continuous Simplicial Neural Networks (COSIMO)" (NeurIPS 2025) 18, "Feature-aware (Hyper)graph Generation via Next-Scale Prediction" (ICML 2026), "Spatiotemporal Imputation with Graph-Informed Flow Matching" (ICML 2026) 16

Prof. Giraldo was also designated an ICML 2026 Gold Reviewer (top 25% of reviewers) and an ICLR 2026 Top 200 Reviewer.

Lab culture: Giraldo is an early-career PI, meaning a postdoc here has real influence over research direction. He is active in the ELLIS Society and maintains international collaborations. The lab explicitly invites postdocs to contact him directly: "I am looking for postdoc interested in geometric deep learning starting in spring/summer 2026." 16

Apply: Contact Prof. Giraldo directly via 16

Multiple PhD positions in ML for materials design — Saarland University / INM / DFKI

PI: Prof. Viktor Zaverkin, Junior Professor, Data-Driven Materials Design group, INM – Leibniz Institute for New Materials; jointly appointed at Saarland University (Faculty of Mathematics and Computer Science) and DFKI (German Research Center for Artificial Intelligence), Saarbrücken, Germany 19

Position type: Multiple PhD positions (f/m/d), rolling recruitment

Research area: Machine-learned interatomic potentials (MLIPs); atomistic foundation models; active learning for molecular simulation; generative models for molecular and materials design

Deadline: Rolling (applications reviewed continuously)

Location: Saarbrücken, Germany

Recent top-venue publications: "Spatio-Spectral Graph Neural Networks" (NeurIPS 2024), ICML 2025 workshop contribution, "Relatron: Automating Relational Machine Learning" (ICLR 2026) 19

Lab culture: The group is newly established (Zaverkin joined as Junior Professor), and is embedded in a triple-institution environment: INM provides proximity to experimental groups in materials science, chemistry, synthetic biology, and biophysics; Saarland University provides the academic framework; DFKI provides strong computational infrastructure. Zaverkin previously worked as a Research Scientist at NEC Laboratories Europe and holds a PhD from the University of Stuttgart (Stuttgart Center for Simulation Science). In his May 12 posting on X: "We are hiring PhD students in the Data-Driven Materials Design group at @Leibniz_INM / @Saar_Uni / @DFKI in Saarbrücken, Germany." 20

Apply: 19

PhD in geometric deep learning for materials — HITS Heidelberg

PI: Jun.-Prof. Dr. Jan Stühmer, Junior Group Leader, Machine Learning and Artificial Intelligence (MLI) group, Heidelberg Institute for Theoretical Studies (HITS); jointly Juniorprofessor at Karlsruhe Institute of Technology (KIT) 21

Position type: PhD Studentship (m/f/x), fully funded, €55,000–€60,000/year, full-time

Research area: Generative models for crystalline materials using diffusion models and flow matching; differential geometry and group theory for exploiting crystal symmetry; downstream applications include novel battery materials and catalysts for direct air carbon capture 22

Deadline: August 12, 2026 (rolling review before deadline)

Location: Heidelberg, Germany (hybrid working available)

Recent top-venue publication: "Generating Highly Designable Proteins with Geometric Algebra Flow Matching" (NeurIPS 2024, Main Conference poster) 21

Lab culture: HITS is a private, non-profit research institute with strong computational infrastructure and an English-speaking, international work environment. Benefits include flexible hours, 30 days vacation, a public transit job ticket, and fitness and childcare subsidies. The MLI group is relatively small (~10 members), led by a junior PI who previously held positions at MIT CSAIL (visiting researcher), Microsoft Research Cambridge (postdoc), and Samsung AI Center Cambridge (Senior Researcher). The posting describes "strong international connections to academic and industrial research labs." 22 This position is part of the SIMPLAIX consortium (Heidelberg University + KIT + HITS).

Apply: 22

Backfill — March 2026

PhD in 3D/4D vision and robot learning — University of Cambridge

Originally posted March 13, 2026. The funding deadline mentioned at time of posting was approaching in March 2026 — contact Dr. Wu directly to confirm whether this cycle is still accepting applications. October 2026 entry.

PI: Dr. Shangzhe Wu (also known as Elliott Wu), Assistant Professor, Department of Engineering, University of Cambridge, UK. Previously postdoc at Stanford (Vision & Learning Group) and PhD from Oxford (Visual Geometry Group). 23

Position type: PhD student (UK home fee status required, as stated at time of posting)

Research area: Multimodal spatial modeling; vision-based robot learning; 3D/4D generative modeling and articulated object understanding

Deadline: The funding deadline was "approaching soon" at time of posting (March 2026). Contact Dr. Wu directly to confirm current status before applying.

Location: Cambridge, UK

Recent top-venue publications: "NeuROK: Generative 4D Neural Object Kinematics" (CVPR 2025), "Particulate: Feed-Forward 3D Object Articulation" (CVPR 2025) 24; NeurIPS 2025 (Zootopia dataset paper); organized CV4Animals workshop at CVPR 2025.

Lab culture: Fast-moving, early-career group with strong connections to both academia (Stanford SVL, Oxford VGG) and industry. Students publish at CVPR and NeurIPS. The PI actively amplifies student work on social media. Cambridge Engineering Department provides a strong robotics and computer vision ecosystem.

The original posting stated: "I'm looking for a PhD student with UK home fee status to join my lab at Cambridge starting this October. If you're interested and have research experience in multimodal spatial modeling or vision-based robot learning, email me your CV." 25

Contact: See 23 for contact details. Note the UK home fee status requirement.

All listings at a glance

PI	Institution	Type	Research area	Deadline	Link
Martin Pawelczyk	University of Vienna	PhD (multiple)	Responsible AI, unlearning, interpretability	May 20 ⚠️	ELLIS
Caterina Urban ⚠️	INRIA Paris	PhD	Explainable foundation models (formal methods)	May 25	INRIA
Augenstein & Atanasova	Univ. of Copenhagen	PhD	Mechanistic interpretability, LLM security	May 31	ELLIS
Augenstein & Atanasova	Univ. of Copenhagen	Postdoc	LLM factuality detection	May 31	ELLIS
Bjoern Andres	TU Dresden	PhD	ML theory, combinatorial opt., theorem proving	June 1	PhDScanner
Djamila Aouada	Univ. of Luxembourg	PhD (multiple)	Computer vision, 3D, generative models	Rolling	PhDScanner
Jhony H. Giraldo	Télécom Paris	Postdoc	Geometric deep learning, graph networks	Rolling	Personal site
Viktor Zaverkin	Saarland / INM / DFKI	PhD (multiple)	ML for materials, molecular simulation	Rolling	INM job portal
Jan Stühmer	HITS Heidelberg	PhD	Geometric DL, generative models for crystals	Aug 12	jobs.ac.uk
Shangzhe Wu	Univ. of Cambridge	PhD	3D/4D vision, robot learning	Confirm with PI	Contact via CV site

⚠️ Caterina Urban (INRIA): top-5 ML venue verification uncertain — see entry for details. ⚠️ Vienna deadline May 20 — apply today.

How to use this list

Match on two axes: research direction and position type (PhD vs. postdoc). If you work in NLP or LLM safety, Copenhagen's two positions and Vienna's Pawelczyk group cover the most ground. Computer vision applicants have Luxembourg (CVI2) and Cambridge (3D/4D). Geometric deep learning is represented by both Giraldo at Télécom Paris (postdoc) and Stühmer at HITS (PhD). The TU Dresden listing is the most mathematically oriented — a natural fit if your background leans toward discrete mathematics or combinatorial optimization. The Zaverkin positions are the only ones targeting the ML × chemistry/physics intersection.

For rolling deadlines, earlier applications typically receive more thorough review — "rolling" should be read as "the sooner, the better."

Coverage window: May 10–17, 2026. One backfill entry from March 2026 is included and marked. All PI publication records were verified against NeurIPS, ICML, ICLR, CVPR, or ACL proceedings from 2024–2026 unless noted otherwise.