Build to understand
I use side projects as engineering laboratories. These projects go below framework-level development into telecom signalling, protocol state machines, distributed Erlang, transformer training, GPU workloads, rendering loops, physics, and reusable systems design.
A hands-on language-model engineering project focused on understanding the full training pipeline by building tokenization, transformer architecture, training loops, evaluation, checkpointing, and scaling experiments from scratch.
PythonPyTorchTransformersBPE TokenizerCUDAMixed PrecisionLLM TrainingA self-contained telecom messaging lab with Go-based ESME, SMPP proxy, SMSC, STP, protocol libraries, fault injection, durable message lifecycle handling, and real M3UA/SCTP support on Linux.
A component-based Erlang/OTP telecom core lab designed around independently buildable network elements, beginning with an advanced STP implementation and an extensible roadmap for MSC/VLR, SMSC, and HLR/AuC components.
Erlang/OTP 29M3UASCTPSS7SIGTRAN
DockerOTP TestingA reusable C++17 2D engine prototype with an ImGui editor, ECS scene model, OpenGL sprite rendering, input mapping, simple physics, prefabs, asset manifests, packaging, and sample games.
I can move from packet formats and transport behavior to stateful application logic, routing, failure handling, and testable network-element boundaries.
I design around concurrency, lifecycle, observability, repeatability, recovery, and clear component ownership rather than only happy-path features.
I deliberately work outside my day-to-day stack, including Go, C++, PyTorch and low-level telecom simulation, to build transferable engineering depth.