This page contains our public Content Delivery Timeline, where you can see the topics we’re actively considering, working on, and planning to deliver — and understand what is likely to be taken up next.
The topics listed here are curated by us based on signals received from multiple channels, including YouTube comments, personal messages/DMs, insights from our corporate trainings, current trends in tech, and recurring pain points observed in practice.
For the topics shown on the timeline, you can register your interest. Higher interest helps us decide which topics to take up sooner, while keeping the process focused and transparent.
Every topic on this timeline is taken to a clear end state: published, rescoped, or discontinued. Topics are not silently dropped.
If you have a topic suggestion that isn’t listed, you’re welcome to share it via YouTube comments (on any video) or by reaching out through Telegram or email. We regularly review these inputs when curating future topics.
1. Physics-Informed Neural Operators for the Poisson Equation
This project demonstrates a Physics-Informed Neural Operator (PINO) approach to the Poisson equation, combining operator learning with explicit enforcement of governing physics. The model learns a global solution operator while preserving physical consistency, bridging the gap between classical PINNs and purely data-driven neural operators.
2. Fast PDE Solution Learning with Fourier Neural Operators
This project showcases the Fourier Neural Operator for learning solution operators of partial differential equations directly in function space. By operating in the spectral domain, the model captures global dependencies efficiently, enabling rapid and scalable prediction of PDE solutions without explicitly enforcing equation residuals during training.
3. Autonomous PINN Single Agent Solver with LLM-Driven Orchestration and Self-Verification
This project demonstrates an autonomous Physics-Informed Neural Network workflow where an LLM generates, trains, evaluates, and iteratively improves a PINN solver end-to-end. The system enforces PDE physics, validates numerical accuracy through tests, and self-corrects until convergence—showcasing agentic AI applied to scientific computing, not just code generation.
4. Multi-Agent Autonomous PINN Solver with Role-Specialized LLM Agents
Build a multi-agent PINN system where distinct LLM agents handle problem specification, solver generation, testing, evaluation, and corrective updates. The workflow forms a closed self-verification loop that enforces physics constraints and numerical accuracy until convergence.
5. Autonomous Code Generation with LLM Agents Using Terminal Bench
This tutorial demonstrates an autonomous code-generation system evaluated using Terminal Bench, where an LLM-driven agent operates directly through a terminal interface to solve real programming tasks end-to-end. The agent plans actions, writes code, executes commands, interprets runtime errors, and iteratively repairs its solution based on terminal feedback—without human intervention. The tutorial focuses on autonomy, tool use, and verification loops, highlighting how agentic systems move beyond static code generation toward executable, self-correcting software workflows.
6. AI-Assisted PINNs Development Workflow using GitHub Copilot (VS Code)
This video demonstrates how GitHub Copilot is used inside VS Code as a development assistant while building a PINN-based PDE solver, focusing on real scientific-computing workflows rather than teaching PINNs from scratch. It shows project structure, code scaffolding, PDE residual and boundary condition implementation, and training loop setup, while clearly highlighting where Copilot speeds up boilerplate and pattern repetition and where human physics judgment and domain understanding remain essential.