Join Astrmira
Bring your judgment.
Build something that matters.
We are looking for people who think independently, create on their own initiative, and work skilfully with AI agents. Research, engineering, and product development meet here: understand the problem, verify the result, and build work that holds up in real use.
Before we begin
Disproving the unit-distance conjecture, cracking the Navier–Stokes problem, forging an empire of trillions of bytes of code…
As AI claims, inch by inch, territories once reserved for genius, we are compelled to ask:
What remains of our meaning as human beings?
An answer is emerging, though most have yet to see it:
Metacognition — the most precious gift nature has given us.
AI does not know what it knows;
nor does it know what it does not know.
It can advance rapidly in a specified direction, yet cannot step beyond the confines of its prompts or examine a problem with the depth of intuition and understanding.
This is not a settled truth. Perhaps, years from now, when AI truly begins to contemplate “Who am I?”, this definition will be overturned.
But before that window closes, we need fellow travellers. In the boundless darkness of the cosmos, supernovae stand for the most dazzling light.
Will you be one of them?
Shared foundations
What every role asks of you.
These are requirements for all three roles. Please consider each one before applying.
Direct agents with intent.
Be an experienced, intensive user of AI agents. Define requirements and acceptance criteria, write explicit documentation contracts that prevent mistakes, and guide an agent toward the outcome you need. Recognise and correct hallucinations quickly. A single prompt followed by uncritical acceptance of the output is not enough.
Know where understanding ends.
Strong metacognition means understanding both what you know and what the agent does not. Step outside the prompt to question the premise, expose missing context, and examine the real problem. Combine your judgment with the agent’s speed, and take responsibility for the result.
Create through your own initiative.
Care about what you will bring into existence. Identify worthwhile problems, make choices, and carry an idea through to a working result. Show us the work you have chosen to build and the decisions you made along the way.
Let the work speak.
Strong publications and open-source experience are welcome, but we assess the substance. A preprint, a niche research direction, or a little-known repository can demonstrate equally strong reasoning and engineering. We look for depth of understanding and your own contribution.
Work well across distance.
Be comfortable with non-traditional workplaces and online collaboration as the default. Make your progress, decisions, and questions understandable to others, and help work move forward across research and engineering.
Respect the act of creating.
You do not need to reinvent every component. Understand existing tools, make deliberate choices, and create something useful from them. Bring curiosity, care, and a clear sense of what your work adds.
Open roles
Three ways to contribute.
The shared foundations apply to every role. Open a role to read its responsibilities and requirements.
Experiments · Analysis · Benchmarks
Data Science Engineer
Turn research questions into credible experiments and evidence.
Experiments · Analysis · Benchmarks
Data Science Engineer
Turn research questions into credible experiments and evidence.
What you will do
- Work closely with researchers to validate and evaluate frontier research, building rigorous and fair benchmarks.
- Use agents to prepare datasets, design experiment frameworks, and shorten research iteration cycles without compromising the quality of the evidence.
What you bring
- A solid foundation in data analysis and statistics; proficiency in Python and the data science stack; skilled use of agents for visualisations and evaluation reports, including clear scientific figures.
- Substantial experience in Experiment, Analysis & Benchmark (E.A.B.) work, with the ability to independently design sound experimental scripts and fair evaluation frameworks.
- Working knowledge of databases, machine learning, and information retrieval; the ability to read advanced mathematical work and translate it into academic and industrial evaluation logic.
- Respect for scientific integrity: identify agent hallucinations and prevent biased comparisons, invalid experimental assumptions, and misleading evaluation results.
Product · Engineering · Agent harnesses
Full-stack Engineer
Carry open-source applications from a working idea to a maintainable system.
Product · Engineering · Agent harnesses
Full-stack Engineer
Carry open-source applications from a working idea to a maintainable system.
What you will do
- Lead the engineering of our open-source applications across the full development lifecycle; use agents to iterate quickly and validate prototypes.
- Write unit tests for modules, collaborate with researchers on new findings and implementation plans, and produce documentation that makes the system easy to understand.
What you bring
- Strong UI/UX judgment, with the ability to reason from first principles about user behaviour, product gaps, and choices beyond the code.
- A thoughtful understanding of agent harness design, including context and prompt management, tool interfaces, long-term memory, and agent-to-agent communication.
- Sound software engineering practices and clear architectural thinking; experience maintaining large repositories is a plus.
- An active interest in open-source communities and technical discussions, especially frontend and UI/UX libraries, with your own judgment about emerging directions.
- Intensive experience using agent products and a practical understanding of different models’ strengths, limitations, and suitability for design or rigorous implementation.
User experience · Integration · Reliability
Product Test Engineer
Find the failures that matter before users have to.
User experience · Integration · Reliability
Product Test Engineer
Find the failures that matter before users have to.
What you will do
- Own end-to-end test design and execution for our open-source applications; build and maintain CI/CD pipelines and realistic integration, boundary, and stress test matrices.
- Use the applications yourself, assess the experience from a user’s perspective, and turn failures into reproducible findings and useful documentation.
What you bring
- Strong testing judgment and hands-on experience with test frameworks, test design, and execution.
- A factual approach and a sharp sense of product boundaries; choose test granularity deliberately and avoid overengineering.
- Guide agents to create meaningful tests that can expose failures, rather than merely restating the implementation, and protect core behaviour through changes.
- Locate anomalies quickly, diagnose their causes accurately, and document the reasoning and reproduction steps clearly.
- Strong metacognition and human-computer interaction thinking: examine the product as an everyday user, developer, and intensive user to uncover overlooked problems.
Additional strengths
Show us what you have made possible.
If any of these apply to you, include supporting material with your application.
- Substantive work in a university or industry lab, with clear personal involvement in the research or engineering.
- Meaningful competition achievements in computing, mathematics, data science, electronic information, or related fields.
- Authorship of strong conference or journal work, or comparable research. Excellent work under review is also welcome.
- Personal open-source projects or substantial community contributions. Repository popularity is not a measure of technical quality.
- Unusual skills or creative work, such as independently designing a game or developing deep expertise in a niche field.
Apply
Start with your work.
Email your résumé or experience summary, representative work, and the role you are interested in. Tell us what you personally contributed and which problems you would like to explore.
What to include
- A résumé or a concise account of your relevant experience.
- Papers, repositories, demos, or other work, with your contribution clearly explained.
- A short introduction to the choices you made, the problems you solved, and why you want to join.