Machine learning engineer for plant DNA (fixed term, April to June 2027) IsoGentiX Ltd Decode:Protect

Society; United Kingdom; Remote; GBP 305,000 - 365,000; Full time

Job summary

IsoGentiX is a young UK company building the genetic record of Madagascar’s endemic plants. We seek UK grant funding to develop a genomic language model that identifies plants from DNA alone and detects genes in alkaloid biosynthesis pathways.

The role focuses on adapting models to long-read data, training dual heads, benchmarking rigorous performance, and documenting results for a technical white paper. This is a 3-month full-time contract, remote within the UK, funded by grants.

Qualifications

• Hands-on experience training a deep learning model on DNA, RNA or protein sequences.
• Experience with contrastive or metric learning and careful benchmark design.
• Proficiency with PyTorch on cloud GPUs and ability to design benchmarks.

Responsibilities

• Adapt an open plant DNA language model to noisy Oxford Nanopore long reads.
• Train two heads: specimen identification and alkaloid pathway gene flags.
• Build and run a benchmark on held-out genomes against current tools.
• Write up the results for a technical white paper.

Skills

Deep learning on DNA/RNA

Contrastive learning

Benchmark design

PyTorch

Education

MSc/PhD in ML or Computational Biology

Tools

PyTorch

Job description

IsoGentiX is a young UK company building the genetic record of Madagascar’s endemic plants, to be sequenced in Madagascar, with community consent and benefit-sharing agreed before anything is collected. We are seeking UK grant funding to build a genomic language model that can tell what a plant is, and which of its genes make medicine-like molecules, from its DNA alone, with no reference genome. Most of the world’s plants have never been sequenced; this is the tool that changes that.

What you would do

• Adapt an open plant DNA language model (such as PlantCaduceus) to noisy, low-coverage Oxford Nanopore long reads
• Train two heads: one that places an unknown specimen (known species, nearest relative, or likely new), one that flags genes in alkaloid biosynthesis pathways
• Build and run a benchmark on held-out genomes against current tools, with targets set in advance
• Write up the results for a technical white paper

You have

• Hands-on experience training (not only using) a deep learning model on DNA, RNA or protein sequence
• Contrastive or metric learning, and careful benchmark design
• PyTorch on cloud GPUs
• An MSc or PhD in machine learning, computational biology or bioinformatics, or equivalent experience
• The right to work in the UK

Nice to have

plant genomics, long-read data, natural products or drug discovery.

Terms

three-month full-time UK employment contract, £30,000 to £40,000 for the three months, remote within the UK. Conditional on grant funding, with a decision expected end of January 2027.

Why this is interesting

data no one else will hold, a real benchmark problem, and a project built to get benefit-sharing right, unlike the Madagascar periwinkle, whose cancer drugs earned over a billion dollars while Madagascar received nothing.