CognitiveXCOGNITIVEX · CAREERS
Build the cognition layer with us
We are a small team building the LCM, the cognitive infrastructure that sits beyond the LLM. If memory architecture, cognitive systems, and product that feels alive are the problems you cannot stop thinking about, we want to hear from you.
WHO WE ARE
A small team with an unreasonable thesis.
Most of the industry is racing to build a better LLM. We think the more interesting frontier is what surrounds the model: a living memory that remembers, reflects, detects patterns, and changes over time. We call it the Large Cognition Model, and the thesis is simple to state and hard to build. The memory is the model.
CognitiveX is the company and the developer platform for that idea. iCog is the consumer app built on top of it, a personal AI with persistent, cross-session memory. We are deliberately small. That means the work is real, the ownership is total, and the distance between an idea and shipping it is measured in days. There is no fabricated org chart here and no committee to route around. There is the problem, the people who care about it, and a high bar.
We are honest about where we are. Some of what the LCM will become is shipped and running today; some of it is vision we are walking toward in the open. If that mix of concrete craft and genuine frontier sounds like the right amount of hard, read on.
WHAT WE BELIEVE
The principles we actually work by.
The LLM is electricity
- intelligence lives in the algorithms, memory, and structure, not the prompt
- the model renders language last; it is the edge of the pipeline, not the core
- if swapping the model changes behavior, the algorithm was never deep enough
Encode structure first
- every module has a real input and output schema before any model touches it
- we can describe what an output should contain without naming a model
- depth over demos: the system should hold up when the lights are on
Small team, high trust
- you own a surface end to end, from the idea to the production deploy
- taste matters: product that feels considered, not just functional
- we hold user memory with care; trust is the product, not a feature
AREAS OF WORK
Where the work is, framed as areas, not reqs.
We are not posting a wall of fake job listings. Instead, here are the problems where another strong person changes what we can build. If one of these is the thing you would do for free on a weekend, that is the signal we are looking for. Titles are flexible; the obsession is not.
- 01
Memory architecture and cognitive systems
The four-tier memory substrate, salience scoring, pattern detection, reflection, and dream consolidation. You think in terms of how memory should decay, consolidate, and surface, and you care about getting the algorithm right before reaching for a model. Research instincts paired with a willingness to ship are the rare combination here.
- 02
Backend and systems engineering
The cognition engine runs on FastAPI and PostgreSQL, serves an SDK, an HTTP API, and an MCP server, and has to be fast, observable, and correct under real traffic. If you enjoy async Python, vector search at scale, careful schema design, and making distributed systems behave, this is your surface.
- 03
Product and frontend engineering
iCog and this platform are where cognition meets a human. We build with React and TypeScript and we care about how the product feels, not just whether it works. You sweat the motion, the type, the latency, and the moment a memory surfaces at exactly the right time.
- 04
Developer experience and platform
The cogx platform is how other builders give their apps living memory. SDK ergonomics, documentation that teaches, MCP integration, and an onboarding that turns curiosity into a working integration in minutes. If great developer tools are an aesthetic to you, this is the work.
WHO THRIVES HERE
The kind of person this is for.
We care more about how you think than about the exact line on your resume. The people who do their best work here tend to share a few traits.
- You are genuinely obsessed with memory, cognition, or the architecture of intelligence, and you have the projects or the questions to prove it.
- You can move between research and shipping. You will read a paper, argue its premise, then write the code that tests whether it holds.
- You take ownership without being asked, you write clearly, and you would rather raise the bar than defend the status quo.
- You are comfortable with the ambiguity of a small team building a new category, and you find that exciting rather than unsettling.
You do not need a specific degree or a particular number of years. If you read the LCM thesis and felt something click, that matters more to us than a checklist.
GET IN TOUCH
No portal. Just write to us.
We do not run an applicant tracking maze. The best way to reach us is a short, specific note. Tell us which area pulls at you, point us at something you have built or written, and say what you would want to work on here. A link to real work beats a polished cover letter every time.
Email careers@cognitivx.io. We read every message and reply to the ones that resonate. If you are not sure you fit a clean role but the thesis grips you, write anyway. Some of the best people do not match a tidy listing.
Want to understand the system before you reach out? That is the best possible move. Read about the LCM, dig into the research, or build on the platform yourself and form your own opinion. The people who show up already fluent in what we are doing tend to be the ones we hire.