We're a high-performing quantitative research group operating in fast-moving, high-stakes environments.
Now we're expanding into
real-time video intelligence
, and we're looking for someone exceptional who can
own and deliver a state-of-the-art temporal vision pipeline
: from large-scale video ingestion and feature extraction to short-horizon event anticipation using advanced transformer architectures.
You'll be working on
cutting-edge temporal action anticipation
, applying transformer-based architectures and multimodal learning on complex, high-speed visual data. This isn't academic work - it's real-world, high-pressure engineering where milliseconds matter.
If words like
context window
,
anticipation window
,
local self-attention
,
temporal grounding
, or
multi-task losses
mean something to you - and you know how to make them
work
- you'll fit right in.
What You'll Do
* Design and train a next-generation model for
short-term event anticipation
in high-speed visual streams.
* Build robust data pipelines and optimize inference for latency and precision - every frame counts.
* Independently experiment with state-of-the-art transformer-based architectures (encoder-decoder, query-based prediction heads, etc.).
* Push models to production-grade reliability with minimal supervision: work autonomously, iterate fast, and deliver results under time pressure.
What We're Looking For
* Deep hands-on experience with
video recognition
/
event anticipation
.
* Strong implementation skills in
PyTorch
,
RegNet / ViT
backbones,
transformer encoders
, and
query-based decoders
.
* Ability to work
end-to-end
* from raw data to working system - under intense iteration cycles.
* A relentless work ethic. We're a small elite team - long hours, weekends, and total focus are standard here.
* Hunger to win and the mindset of a builder, not a researcher.
We're here to take over, not take part.
What You Get
* Work directly with top-tier researchers and traders on one of the most advanced real world prediction systems ever attempted.
* Total ownership of your domain, immediate impact, and the chance to define a new frontier in predictive modeling.
The Environment
Small team. No bureaucracy. Long hours. Competitive compensation. Life-changing upside.
If you understand temporal vision beyond surface-level papers - and can
build it
- we want to talk.
DM or email with examples of work you've done in
short-term action anticipation
or
temporal transformer design
.