Arka
Arka is a dynamic learning-management platform for Konkur candidates: each student's path is modelled as an explicit state machine in the database, the language model sets strategy, and the calendar itself is built entirely by deterministic Python.
The problem it solves
Calendar-based planners break on the first missed day: the backlog piles up on past dates and the student faces a mountain of overdue tasks and a sense of failure. AI planners typically generate a plan once and freeze it, so the moment real life changes, the plan no longer matches it.
The solution
Arka replaces the calendar with a queue — the "conveyor belt". The student is always at a defined position in the curriculum, and a missed day simply means that position is still waiting. The conversational coach holds transactional write access to that same queue: say "I only have 90 minutes today" and the plan is rebuilt immediately, while destructive tools sit behind an explicit two-step confirmation.
A curriculum queue with a daily budget derived from a velocity moving average — overdue work never piles onto past dates.
A WebSocket chat agent with function-calling tools that rewrite the schedule directly; destructive tools require explicit in-chat confirmation.
Every topic runs Learn → Drill → Review → Exam, and the exam score is turned into a mastery verdict by a deterministic state machine.
A visual map of curriculum progress; states are derived at read time with memory-decay awareness rather than stored.
The language model only produces a strategy document and never invents dates; its output passes validation and clamping before anything is stored.
The full app inside Telegram, plus in-app, Telegram and SMS notifications with per-user preferences and quiet hours.
For an organization with a conventional structure, a base rollout takes two to six weeks. Data migration from a previous system and custom modules extend that window.
Yes. Both on-premise and cloud deployment are supported, and the choice does not affect licence pricing.
Yes, through REST APIs and standard web services. For legacy systems a dedicated integration layer is designed.
Under an annual service-level agreement covering business-hours support, defect resolution and periodic version updates.