| Classical Machine Learning |
Neural networks are all the buzz in 2026 and seems rather pivotal to solving problems at scale and I followed up with my project Gradiex.
However, I have also taken the initative to focus on classical machine learning theory and solving for problems with SVMs, decisions trees, KNeighbours, Clustering and other related sub-fields.
Right now my favorite topic is Reinforcement Learning and starting with reward based learning systems : $$ Q(s_t, a_t) \leftarrow Q(s_t, a_t) + \alpha \left[ r_t + \gamma \max_{a} Q(s_{t+1}, a) - Q(s_t, a_t) \right]$$
|
| Programming Languages & Framworks |
A major chunk of softaware and programs written are for developer experience. How do we developing the tooling we require to solve complex problems in constrained environments? With every generaton of
computing technology, we harness new hardware capabilities and existing systems need to adapt and new systems need to be developed. The standardisation of problem solving is essential to progress in
computer science and building better tooling is what I am interested. Developing my own programming language or operating system is not flashy and is difficult indeed, yet it helps me appreciate the progress of
humanity over the last few decades. Focusing on learning low level architectural patterns and kernel interaction : Processes, kernel design, paging, I/O streaming, round-robbin schedular, RAM optimisations.
|