Low-level systems
Modern C++, assembly, concurrency, zero-copy techniques, kernel development, ultra-low-latency paths, and architecture-specific optimization.
COGNISOFT LTD
Bare-metal engineering for ultra-low-latency systems.
Independent contracting in low-level systems, GPU computing, compiler engineering, and AI infrastructure.
ContactAbout
Technical work for latency- and performance-sensitive systems.
COGNISOFT LTD is the contracting vehicle for independent systems engineering work across modern C++, ultra-low-latency software, kernel development, compiler internals, and hardware-aware optimization.
The focus is practical: understand the bottleneck, make the system simpler where possible, and improve the parts where architecture, memory movement, and runtime behavior matter.
Expertise
Areas of work
Accelerators and distributed computing
NVIDIA Hopper and Blackwell, AMD Instinct MI300 and MI350 Series accelerators (MI350P, MI350X, MI355X), Google TPUs, custom kernels, and distributed compute paths.
AI, RLVR, and inference infrastructure
Agentic environments, low-level RLVR environments (reinforcement learning with verifiable rewards), inference optimization, state-of-the-art model serving, vLLM, and ML backends.
Compiler and toolchain
LLVM, backend development, custom tooling, and compiler-level performance work for production systems.
Contact
Discuss a technical problem
For contracting enquiries, send a short note with the system context and the problem you are trying to solve.