Procedure no. DSMBBC/2027/11:Establishing a benchmarking platform for AI/ML predictive models for bioactivity and toxicity in agrochemistry.

Institute of Biochemistry and Biophysics Polish Academy of Sciences; Poland; On-site; PLN 33,000 - 45,000; Full time

Job summary

Institute of Biochemistry and Biophysics, Polish Academy of Sciences, ogłasza dwustopniowy program doktorancki na stanowisko badawcze R1 w dziedzinie chemii i AI/ML. Kandydaci powinni posiadać tytuł magistra w dziedzinie nauk ścisłych oraz doświadczenie w programowaniu i chemoinformatyce.

Program zaczyna się 1 marca 2027 r. Projekt koncentruje się na budowie benchmarków ML dla agrochemii i ocenie modeli predykcyjnych z zastosowaniem danych eksperymentalnych, z możliwością uczestnictwa w

Qualifications

• Preferowane doświadczenie w programowaniu, chemii obliczeniowej i umiejętności z zakresu cheminformatyki oraz ML.
• Języki: angielski na poziomie Excellent.
• Dziedzina badań: chemia.

Skills

Programowanie

Wiedza chemia obliczeniowa

Cheminformatyka

Uczenie maszynowe

Education

Magister nauk chemicznych

Magister inżynier

Lekarz medycyny (lub równoważny)

Job description

Institute of Biochemistry and Biophysics Polish Academy of Sciences

Organisation/Company Institute of Biochemistry and Biophysics Polish Academy of Sciences Research Field Chemistry Researcher Profile First Stage Researcher (R1) Positions PhD Positions Application Deadline 30 Nov 2026 - 23:59 (Europe/Warsaw) Country Poland Type of Contract Temporary Job Status Full-time Hours Per Week 40 Offer Starting Date 1 Mar 2027 Is the job funded through the EU Research Framework Programme? Not funded by a EU programme Is the Job related to staff position within a Research Infrastructure? No

Offer Description

Modern crop protection faces unprecedented challenges, mainly in the form of pests and weeds evolving resistance, and the lack of new modes of action (MoA) reaching the market. Further, climate change shifts pest pressure and geographical distribution, while regulators and society increase the demand from regulators and society for rapid reductions in both chemical load and off-target toxicity
Fair evaluation of machine learning (ML) models is a necessary task of basic research in artificial intelligence. It consists of designing high-quality datasets, challenging train-test splits, and appropriate evaluation metrics and procedures, in order to obtain a good estimation of generalization capabilities of ML models. Despite AI/ML advances in pharmaceutical chemistry, comparable progress hasn't reached agrochemistry. Scarcity of QSAR models are calibrated on pesticide-relevant chemical space, while non-expert researchers cannot determine whether sophisticated graph neural networks outperform classical random-forest baselines for agrochemical discovery, nor gauge how much experimental data are required to reach regulatory accuracy thresholds. By systematically assembling high-quality experimental records and constructing rigorously validated benchmark tasks mirroring real-world regulatory questions, this programme addresses principal bottlenecks in digital agrochemical innovation. It integrates data engineering, algorithm development, together with domain knowledge and experimental validation forming a feedback loop in order to deliver its results.
This programme will deliver a task-oriented benchmark analogous to ImageNet, MoleculeNet, or TDC for crop-protection chemistry. Coupled with transparent baseline models and public leaderboard, we will catalyse comparable, trustworthy AI in agrochemistry. A mature benchmarking ecosystem will impact: (i) algorithmic research focusing on realistic agrochemical challenges, (ii) reduced redundant wet-lab screening by prioritising promising scaffolds, and (iii) evidence-based regulation by quantifying in silico prediction uncertainty. Our goal is to drive impactful and reliable advances in machine-learning research, unlocking ML’s potential and impacting allied domains like veterinary parasitology and environmental toxicology, enhancing academic and societal benefits.
The proposed programme builds on substantial prior experience considering generating data assets, modeling protocols, predictive ML architectures and experimental validations in agrochemical research. CheMM Lab has recently developed ApisTox, the largest dataset on small-molecule pesticide toxicity toward honey bees. They have comprehensively evaluated diverse machine learning approaches for predicting pesticide toxicity to honey bees, including molecular fingerprints, graph kernels, graph neural networks, pretrained transformers, and explainable ML approaches.
The programme is structured with two main research questions:
1) How does the ranking performance of state-of-the-art molecular AI/ML models change when they are evaluated on agrochemical rather than pharmaceutical features and endpoints?
2) Can a chemically diverse, less pharma-biased pre-training corpus improve the generalization of neural models across distinct agrochemical tasks?

Hold a degree of Master of Science [Magister], Master of Engineering [Magister Inżynier], medical doctor or equivalent in the field of: exact sciences, natural sciences, medical sciences or related disciplines, granted by a Polish or foreign university; a person who does not possess the qualifications described above may take part in the competition, but must obtain the qualifications in question and provide the relevant documents before the start of the programme at the Doctoral School (i.e., 1st March 2027). Education at the Doctoral School begins on 1st March 2027

Specific Requirements

Preferably, candidates have experience in programming, knowledge of computational chemistry and possess cheminformatics together with machine learning skills.

Languages ENGLISH Level Excellent

Research Field Chemistry

Additional Information

• Possibility of renting a room in the Institute's hotel for the first 3 months at a special price (depending on rooms’ availability);
• Possibility to join a medical plan;
• Research mini-grants for doctoral students (up to PLN 20000, a PhD student may apply for funds twice during his/her studies);
• Conference Grants (each PhD student can use up to PLN 12,000 to attend conferences during his/her studies);
• Surcharge for the multi sport card;
• Support in obtaining a temporary residence card (+ Reimbursement of TRC application expenses);
• Language courses (Polish, English, German, French, Spanish, Italian… (the offer of courses may change);
• Psychological care (up to 5 visits);
• Subsidy for the purchase of eyeglasses (once every two years);
• Free parking spot.

Selection process

The recruitment process consists of three stages:

1. selection of candidates by the Committee based on the alignment of the candidate’s competencies with the requirements specified in the recruitment program;
2. selection of candidates by the Committee based on their previous achievements and academic results presented in the submitted documents;
3. an interview conducted by the Committee including in particular:
   • a presentation delivered by the candidate containing the outcomes of his/her research (a Master’s thesis or other research work carried out by the candidate); the presentation must not last longer than 10 minutes;
   • questions asked by the members of the Committee related to the presented project, the methods used and interpretation of the results obtained;
   • questions asked by the members of the Committee related to the proposed PhD programme described in the recruitment announcement;
   • questions related to the candidate’s motivation for scientific work.

Stage One

The Committee analyses the application documents in terms of the candidate’s competencies and their alignment with the requirements specified in the recruitment program. Candidates who obtain at least 40% of the maximum number of points in the first stage qualify for the second stage.

Evaluation criteria:

• education in the subject areas listed in the application guidelines; (scoring scale: 0.0 – 2.0 points);
• experience in the subject areas listed in the recruitment program (scoring scale: 0.0 – 2.0 points);
• other competencies specified in the recruitment program (scoring scale: 0.0 – 1.0 points).

Stage Two

The Committee evaluates the candidates’ previous achievements and their academic performance based on the submitted documents. Candidates with the 5 best scores, not lower than 60% of the maximum number of points, will qualify for the third stage;

• results obtained during studies (scoring scale: 0.0 – 4.0 points)
• competencies specified in the recruitment programme (scoring scale: 0.0 – 3.0 points)
• co-authorship of a scientific publication (depending on the role in the publication) (scoring scale: 0.0 –1.0 point)
• other achievements, e.g., awards, active conference attendance, honourable mentions, scholarships, domestic and foreign internships, voluntary work, popularisation of science (scoring scale: 0.0 – 2.0 points)

Stage Three

When determining the results of the interview, the Committee takes into account:

• the degree of understanding of one’s own work; the ability to interpret the results obtained; the understanding of the methods used, knowledge in the field related to the topic of the submitted work (a scale of the evaluation: 0 - 10 points);
• knowledge in the subject matter listed in the recruitment announcement (a scale of the evaluation: 0 - 10 points);
• form of presentation of the candidate’s results (a scale of the evaluation: 0 - 3 points).

Number of offers available 1 Company/Institute Institute of Biochemistry and Biophysics Polish Academy of Sciences Country Poland City Warsaw Postal Code 02-106 Street ul. Pawińskiego 5a Geofield