The inventors building
what comes next.

A community of researchers at the frontier. Your published work is your proof of capability. We help the right deep-tech teams find you.

US Patent App · 2024/0312847 A1
A system and method for real-time anomaly detection in distributed sensor arrays using sparse autoencoder architectures. Filed: Mar 2024. Inventors: Chen S., Osei A.
→ Anomaly-detection systems
doi:10.1038/s41586-024-07218
Extended Data Fig. 4 — Binding affinity comparison across 847 protein complexes in the holdout validation set.
→ Protein-binding analysis
github.com/k-nakamura/qec-surface · README
# Surface Code Simulator Fast, hardware-accurate QEC simulation. ★ 2.4k · MIT License > pip install qec-surface
→ Open-source research software
Algorithm 1 — Graph Encoder
Input: G = (V, E), dim d, depth L Output: h ∈ ℝ^(|V|×d) 1: for l = 1 to L do 2: aggregate neighbourhood
→ Graph representation learning
SBIR Phase II · UKRI · EP/X024891/1
Milestone 3 Complete — 94% capacity retention over 500 charge cycles in solid-state prototype. Phase II award confirmed: £750k.
→ Prototype-to-milestone delivery
Fig. 3 — Ablation Study
Removing the cross-attention layer yields a 7.2% performance drop, confirming its role in cross-modal feature alignment.
→ Cross-modal architecture design
Acknowledgements
Supported by EPSRC EP/T026170/1 and a Royal Society University Research Fellowship (URF\R1\211218). Computing via ARCHER2.
→ Independently funded research
Abstract
We present a scalable framework for learning transferable representations from heterogeneous scientific corpora across research domains.
→ Transferable representation learning
Hover a fragment — see what Asriel reads
The process

Three steps.

01
Connect your ORCID iD

One click. Your ORCID iD links directly to your verified publication record — no manual entry, no forms.

02
Asriel reads your work

Asriel reads your published work and builds a deep capability profile — what you've built, what domains you've pushed forward, what problems you can actually solve.

03
The right teams reach out

When a role matches your research depth, companies see you — ranked by fit, not by who submitted an application first.

The problem

Your work deserves better signal.

01
Your CV doesn't show what you can build

Years of building genuine expertise compressed into job titles and keywords. The depth of what you actually did, the problems you solved, the systems you built — none of it survives the format.

02
Your publications don't show who's looking

Google Scholar tells the world what you've published. It doesn't tell the deep tech team in Boston that they need exactly your work on sparse attention mechanisms, right now, this quarter.

03
Cold applications don't reward depth

The companies you'd thrive at are drowning in applications. Yours gets ranked by keyword match and submission speed, not by whether your research actually solves their problem.

The community

What our community works on.

A sample of the research being done across the platform. Open any card to see the capability profile Asriel built from their papers.

Sample profiles — illustrative of the community. Every profile is built the same way: from the papers, not the CV.

Your work deserves
to be found.

Connect your ORCID iD in one click. Free forever.

Free foreverPasswordless