The Convergene
Self-improving API optimisation
Every team was spending hours tuning temperature, model, and token settings or just sticking with defaults.
We asked ourselves: what if optimisation could optimise itself?
That hackathon project ended up winning the Reinforcement Learning Track at Weavehacks-2 (W&B + Google Cloud).
Now it’s grown into The Convergence: an open-source framework that automates API optimisation using evolutionary algorithms, multi-armed bandits, and reinforcement learning.
It learns what works, remembers across runs, and balances cost, quality, and latency automatically for AI performance.
We open-sourced it because every AI team faces this problem, and the infrastructure for continuous, transparent optimisation should be shared.
If you’re tired of guess-and-check prompt tuning, come explore how we’re making AI systems self-improving by design.
🏆 Devpost: https://devpost.com/software/the-convergence#updates
💻 GitHub: https://github.com/persist-os/the-convergence
1 Comment
This is fantastic! The Convergence is a godsend for AI engineers. Manually adjusting parameters is too time-consuming, and automated optimization is definitely an efficiency booster. The open-source spirit is also commendable! 👏