This is a growth hire, adding to Climate Policy Radar’s data science capacity. Reporting to the Head of Data Science, you'll work closely with our existing team of data scientists to analyse our datasets, build tools to deepen that analysis, and then productionise those tools for users to leverage in their own research.

As a skilled generalist within NLP, you'll work on a range of interesting problems here, with a range of interesting people! You'll collaborate with climate experts to understand the problems we need to solve, and with software engineers to take your work from research through to production.

About the data science team

Climate Policy Radar's engineering org is split into a few functional teams: programmes (domain experts), platform (high-quality data sharing), application (user-facing tools), and data science (models and evaluation). Despite those splits, most of our work is cross-functional.

The data science team builds the models that power our search engine, classifiers, and LLM workflows, and the evaluation frameworks that keep them honest. Our work directly informs policy decisions, so we care a lot about evaluation, monitoring, and minimising bias. We're research-informed but production-focused, and we default to working in the open by publishing datasets, models, and papers wherever we can. We also share what we've learned through blogs, papers, and public talks.

Tech preferences

The vast majority of the data science team’s work is written in Python, so we’ll expect you to be a fluent reader/writer!

In addition, here’s an (unordered) list of tools which we work with regularly. We’re always open to new suggestions of tools you've had success with, but familiarity with these tools will help you integrate with our existing stack:

• ML & NLP: PyTorch, Huggingface Transformers, Pydantic AI, Pandas, Spacy

• APIs & backend: Python, FastAPI, Pydantic

• Data & infrastructure: DuckDB, Snowflake, PostgreSQL, Docker, AWS (ECS, S3, etc.), Pulumi, Prefect, GitHub Actions

• Search & demos: Vespa, Streamlit

• Testing, evaluation & monitoring: Pytest, Hypothesis, Weights & Biases, Posthog

• Development tools: GitHub, Cursor, Claude Code

What will you be working on?

Climate Policy Radar's core product is a search engine for 30,000+ climate documents. That engine is supported by a set of single-class classifiers which detect mentions of important climate concepts (e.g. flooding, or deforestation). In your role, you're likely to:

• Work closely with the programmes and platform teams to help shape our research & development priorities

• Develop, evaluate, and deploy state-of-the-art NLP models which extract structured information from climate-related documents, with an emphasis on small, efficient architectures

• Help to build and deploy APIs which demonstrate new model capabilities, for other teams to prototype around

• Contribute to building evaluation harnesses for search relevance, classifier accuracy, data quality, etc, which allow other teams to reliably and independently iterate on work

• Implement approaches for data labelling or sampling to more efficiently create training and evaluation datasets

• Analyse real user search logs and behavioural analytics to describe their journeys across our site and understand their needs.

• Research using LLMs to automate tedious parts of the climate policy research process, and partner with experts to build evaluation datasets which ensure that we retain our high standards for accuracy and bias

To hit the ground running in this role, we anticipate that you'll have the following:

• 3-5 years experience tackling applied NLP problems with machine learning

• Demonstrated ability to take ideas from proof-of-concept to working implementations, in partnership with other teams

• Excellent ability to write well tested, maintainable, version-controlled Python code

• Strong working knowledge of applied machine learning techniques (both supervised and unsupervised). We'd expect you to understand how to work with these models effectively to obtain good results

Nice to haves

In addition to the above, we’d love to hear from you if you’ve got experience with any/all of these:

• Applying LLMs to highly structured or technical datasets

• Identifying and mitigating cultural biases in NLP models/datasets

• An emphasis on building small, task-focused models - distillation, transfer learning, and pre-trained model fine-tuning are all topics that we care about

• Techniques such as active learning, weak supervision, and zero/few shot learning in NLP

• Deploying models to production systems, with consideration to what makes an effective user-facing model

• Building and/or analysing knowledge graphs

• Developing around the Wikidata / Wikibase / Wikimedia ecosystem

• Search relevance evaluation and optimisation

We are looking for candidates with significant experience in highly collaborative cross-functional teams, excitement about working in a startup/scaleup environment and all that brings. 

We are a mission driven organisation, and work best with people who have strong alignment with our values. We care about them deeply.

We actively encourage applicants from diverse and historically underrepresented backgrounds. Not sure if you tick all the boxes but feel like you align with our values, are excited about working in Climate Change and AI and have the potential to do well in the role? Click apply! We’d love to hear from you.

• Salary ranging from £60k - 75k pa, depending on experience

• A deep commitment to employee wellbeing, including policies such as 4 day workweek (same pay, Fridays off), unlimited annual leave, and a wellbeing allowance

• A vibrant, collaborative, empathetic work culture that thrives on innovation and the impact of our work

• Hybrid work environment (2 days a week) in London, currently at Sustainable Ventures in County Hall with a plan to move to TechSpace in Goswell Road in late March. We welcome conversations about flexibility and remote work to accommodate individual needs

We know that applying for a new job can be full of uncertainties - and we aim to reduce those by communicating clearly. Our process is made of several stages (see below). After each stage, we’ll contact you as soon as we can and no longer than 2 working days, to let you know if you will be progressing to the next stage.

If you have a disability or special need that requires accommodation during the process of application and selection, please let us know.

Process:

1. 30-60 minute screening call with our recruiter

2. 60 minute remote behavioural interview with two team members,

3. Task:

   1. 2-3 hour take-home task (scheduled to be sent and returned at a time that suits you)

   2. 60 minute interview to discuss your task

4. 30 minute in-person final fit interview with CEO and Head of People (you’ll also have the  opportunity to meet the team at this stage - not a part of the interview, but gives you the opportunity to get to know the team and learn more about us in an informal setting)

5. Offer subject to references

We’ve all felt the anxiety of waiting to hear back from interviews, so we will contact you no later than 2 working days after each interview to let you know if you will be progressing.

Unfortunately, we are currently unable to sponsor work visas. Only applicants legally authorised to work in the UK will be considered.

At Climate Policy Radar, We are committed to fostering a workplace that is inclusive and equitable. Climate Policy Radar welcomes applicants from all backgrounds and does not tolerate discrimination in any aspect of employment. We actively work to ensure equal opportunities for all, regardless of heritage, ancestry, national origin, citizenship, religion, sex, sexual orientation, gender identity, age, disability, relationship choices, or criminal history, in line with legal requirements. We also consider qualified applicants regardless of criminal histories, in line with legal requirements.

Not sure if you tick all the boxes but feel like you align with our values, are excited about working in climate change and AI and have the potential to do well in the role? Click apply! We’d love to hear from you.

If you have a disability or special need that requires accommodation in the process of application and selection, please let us know. 

If you have a disability or special need that requires accommodation in the process of application and selection, please let us know.