Lossfunk is a lab for independent researchers, not a startup or a commercial enterprise. It’s a place where like-minded people come together to explore foundational questions in AI and beyond. Think of Lossfunk as more like the Santa Fe Institute than OpenAI: we’re not building foundation models or products; we’re a research lab focused on understanding how intelligent systems and complex phenomena work.
What you get at Lossfunk
Growth philosophy
We see the lab as a launching pad for talent. People are typically associated with us for a short, intense period (typically 6-24 months), do great work and then go on to launch their careers in industry, graduate school, or founding a startup. We help people realise their career dreams and celebrate when they move on.
Financial growth
Full time researchers (interns and fellows) get a stipend while they’re associated with the lab. This stipend is reviewed once a year and we aim to adjust it in line with inflation / cost-of-living changes. Lossfunk is not the right place if your main goal is rapid salary growth; we prioritise freedom, ideas, and long-term career capital over compensation maximisation.
Since Lossfunk doesn’t have commercial ambitions and is currently funded by my personal funds (Paras Chopra), Lossfunk will not compete with other commercial labs on talent or compensation. We don’t intend to raise VC funding or pivot into commercial-oriented research, so there are no ESOPs at the lab.
Spinouts (rare and optional)
The lab is focused on foundational questions, not applications. In the rare case that an opportunity to commercialize research arises, our approach will be similar to universities: we spin out a separate entity. When that happens, the founding team will receive the majority of the equity, people who helped in the research may receive advisor equity and Lossfunk will keep a small share (5-10%) as a licensing fee. A spinout requires an external pre-seed/seed commitment as validation.
Spinouts are a possible side-effect of good research, not the goal of the lab. You should join Lossfunk for the work and the environment, not in expectation of a future startup. Most projects will never become companies, and that’s by design.
At Lossfunk, we believe great research comes from curiosity, openness, rigor, reciprocity, and playfulness.
You may want to go through following resources to align yourself with how we think about science and research:
We prefer asynchronous communication and documentation, so Slack is the beating heart of Lossfunk.
Once you join Lossfunk, you will be given access to our Slack. Once you get on the Slack, note:
Once you’re ready to kickstart your project, here’s what you need to do:
We strongly encourage daily updates on your channel. You can write about what you did, learned, or post interim results. Doing this is extremely vital as your updates are the only way to invite attention, collaboration and feedback from other people at Lossfunk.
Daily or near-daily updates are encouraged to maintain project momentum and visibility. Short summaries or even brief reflections are sufficient. Daily updates should take no more than 10 minutes and are often a few bullet points (with appropriate links).
Occasional misses are fine, but if you fail to keep your project channel updated for a long period of time (>two weeks), we will assume the project is inactive, and could be liable to get deleted/archived from Slack.
We have weekly synchronous meetings where the Lossfunk founder and other researchers come together to discuss ongoing projects and brainstorm on big open questions (such as those listed on our website).
We look at the following criteria in order to decide who gets included in these sync meetings and hence our special attention:
What special attention means for projects:
Once your project is invited for weekly synchronous meetings:
Your project is liable to be removed from weekly meetings (hence special attention) once:
So, in a nutshell, Lossfunk is a reciprocal relationship. You get more from Lossfunk if you put more into it.
We follow usual conventions from academia:
In short, anyone with substantial intellectual contribution to a project should be added as a co-author. The primary author has final responsibility for determining authorship order and inclusion, but this should reflect transparent discussion among contributors. If you’re not clear, err on the side of asking people if they wish to be considered as a co-author.
Note that authorship comes with responsibilities. So, all authors on a publication are expected to:
You can refuse to be a co-author on a paper if you’re unable or unwilling to take on these responsibilities.
In case of unresolved authorship disputes, an internal committee (including at least one uninvolved senior researcher) will mediate.
In exceptional cases, we may require code not to be made public (for safety, operations or commercial reasons). In these cases, Lossfunk team will reach out to you.
We provide GPUs, compute and OpenRouter credits for active projects. Moreover, we also provide 1-1 discussion opportunities with Lossfunk founder and senior researchers.
To avail these resources, please contact EA team at [email protected] or message in #lossfunk-support slack.
Resources are allocated based on project activity and alignment with Lossfunk focus areas. Allocation may change as projects evolve.
At Lossfunk, we want creativity to thrive while keeping IP ownership transparent and fair. The general principle is simple:
In all cases, authorship and moral credit remain with individual contributors.
Research projects typically take 6-9 months to complete (from conception to paper release). You’re welcome to kickstart a new project at that point by making a fresh research plan.
However, at all times, to remain part of Lossfunk, you’re expected to remain active on Slack and contribute to discussions. To keep the community active, inactive Slack accounts are periodically reviewed. You can always rejoin by re-expressing interest.
We support travel for the research that reaches major peer-reviewed milestones. To encourage high-rigour work, we sponsor travel and conference attendance under the following conditions:
A* Main Conferences:
If a paper is accepted at an A* main track venue (see the list below), Lossfunk will fully sponsor travel and attendance. (Note: Workshops are not included; we treat them as spaces for feedback, iteration, and early-stage ideas.)
Top-Tier Science Journals:
If work is accepted in a major science journal such as Nature, Science, or an equivalently selective journal, we will sponsor travel or attendance for a relevant national or international meeting.
Who is eligible for travel sponsorship
Primary author with Lossfunk as their affiliation.
Guiding Philosophy:
Workshops are valuable for discussion and early refinement. But our goal as a lab is to publish in high-rigour main conferences or major journals. Travel sponsorship is designed to reward and celebrate reaching that level. Case to case exceptions will be made based on these guiding principles.
A* Canon:
Lossfunk recognises the following as top-tier A* / main-track venues: ICML, ICLR, NeurIPS, AAAI, IJCAI (core ML/AI); CVPR, ICCV, ECCV (computer vision); ACL, COLM, EMNLP, NAACL (NLP); and RSS, ICRA, IROS (robotics). These are the conferences for which Lossfunk sponsors travel upon acceptance.
A* ML Journals:
Lossfunk recognises the following as top-tier machine-learning journals for publication-based travel sponsorship: JMLR, Machine Learning (MLJ), Transactions on Machine Learning Research (TMLR), IEEE TPAMI, and IJCV; and for broader scientific impact, journals such as Nature, Science, PNAS, Nature Machine Intelligence, and Nature Communications.