AI research intern: Molecular Design Pipeline

About

Biolevate builds AI-powered software that turns unstructured scientific data into reliable, collaborative workflows for medical writing and regulatory work. Our platform helps scientists, clinicians and medical writers move faster while maintaining the rigor their work demands.

Founded in Paris in 2024, we combine a commercial SaaS platform with in-house research in oncology and dermatology. It is an unusually practical place to work on AI: the systems we build are used both to advance scientific research and to help bring therapies to patients faster.

Biolevate Science team

At Biolevate, the Science team is building a new way to think about biology. We start from a simple conviction: disease is not random chaos, but a biological language - a logic of how life adapts, resists, and reorganises under pressure. Our scientists use advanced multi-omics reasoning to listen to this language across genes, RNAs, proteins, pathways, and clinical context, and to reconstruct the rules that drive health and disease.

Rather than generating ever more lists of genes, we work to integrate diverse biological layers into coherent programs and convert them into testable, traceable hypotheses. This helps us identify mechanistic patient subgroups, uncover hidden regulators of response and resistance, and prioritise therapeutic targets with a clear biological rationale. Along the way, we rely on platforms like OCEAN to support a more explicit and structured way of reasoning through complex biology.

The goal is always to close the loop: we design experiments, validate hypotheses in cellular and ex vivo models, and translate what we learn into patents, biomarkers, new targets, and smarter trial strategies. In practice, this means moving from data to decisions - using evidence that is mechanistically grounded, auditable , and strong enough to support real R&D choices.

Within a few years, this approach has already contributed to a growing patent portfolio, redefined disease logic in oncology, and uncovered patient categories invisible to genetics alone. At Biolevate, scientists remain in control of meaning while AI expands their capacity to explore, explain, and anticipate biology’s logic. Together, we are building a shared intelligence of biology - one that connects discovery to real clinical impact for patients.

You will contribute to the development of an AI-driven chemistry pipeline for therapeutic molecule design, optimization, and synthesis planning . The objective is to build a computational framework capable of proposing candidate molecules for prioritized biological targets, evaluating their drug-relevant properties, and assessing their chemical and synthetic feasibility through retrosynthetic analysis. The project will explore recent advances in molecular machine learning, chemistry-focused language models, graph-based methods, and structure-aware approaches .

Depending on the performance and limitations of existing open-source models, the work may involve their integration, benchmarking, fine-tuning, or the development of complementary models and representations. The broader goal is to extend our existing multi-omics target discovery pipeline by connecting biological target identification with downstream computational chemistry and molecular design. This would establish a more integrated workflow linking target discovery

• molecular generation and optimization
• pharmacological assessment
• synthesis planning . A key aspect of the internship will be to determine how different AI approaches can be combined into a robust and practically useful pipeline, rather than considering molecular generation as an isolated task. Generated candidates should therefore be evaluated not only for their predicted relevance to a biological target, but also for their physicochemical and pharmacological properties, structural plausibility, and potential synthetic accessibility. Internship goals
• Explore state-of-the-art AI methods for chemistry , including chemistry-focused LLMs, molecular foundation models, graph neural networks, and structure-aware approaches. Identify the most appropriate models and representations for molecular generation, optimization, reaction modeling, and synthesis planning, and characterize their strengths and limitations for the intended drug-discovery workflow.
• Develop a molecular generation and transformation system capable of proposing novel compounds or optimizing existing chemical structures for prioritized biological targets. The system will investigate molecular representations including SMILES, molecular graphs, and, where relevant, 3D structural representations , with the aim of enabling controlled molecular modifications and optimization toward defined biological and chemical objectives. This step may involve use of SOTA methods or, if any flaw or important limitations has been identified, development of novel methodologies.
• Integrate pharmacological and developability assessment into the candidate-

Generated molecules will be evaluated using relevant physicochemical, drug-likeness, and ADMET predictions , allowing unsuitable candidates to be filtered and promising compounds to be prioritized according to multiple criteria rather than molecular-generation scores alone.

• Develop a retrosynthesis and chemical-planning module to assess whether proposed molecules can realistically be synthesized. The pipeline will investigate existing reaction-prediction and retrosynthesis approaches to generate candidate synthetic routes, evaluate synthetic accessibility, and, where relevant, support reaction prediction or optimization. The resulting system should provide a simple interface connecting prioritized biological targets
• candidate molecules
• predicted properties
• proposed synthetic routes . The final pipeline will be demonstrated on a concrete internal research use case , using a biological target identified through our multi-omics analyses to generate or optimize candidate molecules and assess their predicted pharmacological properties and synthetic feasibility.

Why Biolevate

• Meaningful mission: We are building a new way to understand biology - one that treats disease as a causal, mechanistic system rather than a collection of disconnected signals. As a researcher at Biolevate, your work helps turn complex biological data into clearer hypotheses, sharper questions, and more relevant scientific insight.
• Scientific impact: Your research will contribute directly to the discovery of mechanistic patient subgroups, actionable targets, biomarkers, and smarter trial strategies. The work is not theoretical for its own sake: it is designed to move from biological signal to meaningful scientific and clinical decisions.
• Deep biological reasoning: You will work with a platform built to integrate multi-layer evidence across omics, pathways, network logic, literature, and pharmacology. That means less time spent navigating fragmented data and more time uncovering true drivers, hidden regulators, resistance mechanisms, and the biological context that gives results meaning.
• Rigor and traceability: Biolevate is a place for researchers who care about evidence quality. We value explicit reasoning, ranked evidence, auditable logic, and testable hypotheses - so the science is not only insightful, but defensible and reproducible.
• Close collaboration and partnership: Research here does not stop at analysis. We work in close partnership with internal teams and external collaborators to turn biological insight into shared next steps, aligned interpretation, and stronger scientific decisions. That creates a true loop between computation, biology, and collective problem-solving.
• Ownership and growth: You will have the opportunity to shape how new biology is discovered and interpreted, with room to influence methods, standards, and scientific direction as the company grows.
• Paris-based deeptech environment: Join a team working at the intersection of biology, data, and decision intelligence in a strong French deeptech ecosystem - with the ambition to create tools and insights that matter globally.

Biolevate

We value authenticity and benevolence . Every team member contributes their unique background and life journey, which makes us a stronger collective. Beyond work, we genuinely enjoy spending time together, team dinners, weekend treks, sport sessions, offsites, and many shared experiences that build strong bonds.

At Biolevate, you will be part of a dynamic, international team shaping the future of drug discovery and medical innovation. We combine deep science, cutting-edge AI, and internal research to deliver real-world impact. Join us to build a global brand with purpose and grow in a high-trust, high-growth environment with a compelling package that includes stock options.

• 💼 Attractive Financial Package: We offer a compelling package, which includes a competitive salary and stock options, that recognizes your skills and experience, ensuring you feel valued for your contributions to our innovative projects.
• 📍Private Office: Station F , the world’s largest startup campus, in Paris.
• 🏡 Flexible Work Arrangements: We support work-life balance with flexible working hours and options for remote or hybrid work arrangements, understanding the value of flexibility in today's dynamic world.
• 🌎 Collaborative & International Culture: Experience a positive, inclusive, and collaborative work environment. We value each team member’s contribution and encourage sharing ideas and feedback.
• 🤝 Authentic Team Culture: We value authenticity and benevolence. Our diverse background make us stronger, and we enjoy building real connections through team moments like dinners, sport sessions, offsites, and shared experiences.
• 🍽️ 💻 Other perks : Meal vouchers, Latest Macbook and tech

Interview Process

can be adapted when needed, but generally includes the following steps :

• First screening interview (visio)
• Meet the Biolevate team at Station F and meet founders
• Internal review
• Offer