An engineering team in Pune.
FeynCode started in 2023 with three engineers. We spent time learning how our clients worked before building their software. That work led to Kaeonix, which we now develop alongside client projects.
How we got here
We spent the first two years on client projects: an HR platform for an agency managing work through group chats, production tracking for a manufacturer and invoicing for a food trader. Across these businesses, people spent a lot of time moving information between systems.
We began working on Avatar to help with our own software work: reading repositories, proposing changes and running tests. What started as an internal tool became the coding model behind Kaeonix.
We develop Kaeonix and take on client projects where we can apply what we learn. Those projects also help us understand how agents work in businesses with daily responsibilities and deadlines.
How we work
We work in teams of two to four. Each engineer understands the wider project, and the person who will maintain the software takes part in the decisions.
We document the architecture, keep a readable changelog and write operating guides that are easy to follow under pressure. Your team should be able to use them after our work is finished.
Sometimes a problem does not need new software, and we will tell you when that is the case. Two of our last ten enquiries ended with a two-page note and no invoice.
What you can expect from us
If your project uses a model, we test it on your data before designing the interface. Knowing what it gets right and where it struggles helps us design the product and avoid reworking it later.
You own the source code, infrastructure and documentation, along with any model weights we train for you. Everything lives in your accounts from day one, so you can keep working without depending on us.
Every engineer understands how the whole project fits together. That reduces handover work and helps us give realistic estimates. When you have a question, you can speak to the person who wrote the code.
We keep the foundations simple: one database, one queue and one cache, using tools we know how to debug under pressure. We try new approaches where they offer a useful improvement.
Interested in joining the team?
We hire twice a year and take time over each decision. Send us something you have built, along with a short paragraph about what you would change if you built it again.