Software Engineer · AI & Systems
I build software across AI, systems, developer tooling, automation, and web products.
My work focuses on understanding how systems behave, connecting technologies into useful tools, and validating what actually works.
| Project | What it explores |
|---|---|
| LapisLLM | An experimental language-model engineering project: tokenizer, Transformer components, training workflows, evaluation, checkpointing, and inference. |
| CHAD | An AI application/runtime project in the LapisLLM ecosystem, with conversation handling, model integration, persistence, and a CLI. |
| NETWATCH | Linux network visibility and authorized administration tooling, including discovery, diagnostics, traffic management, and structured exports. |
| Vibe Coding Instructions | A repository-native engineering workflow for planning, implementation, security review, testing, and evidence-based verification. |
| Thread Bot | An automated, source-grounded content workflow with deduplication, quality checks, scheduled runs, and reporting. |
| Handshake | An Android-first prototype exploring trusted-contact workflows and advisory analysis of potentially coercive messages. |
| myportfolio | A static portfolio documenting selected web work, project context, implementation decisions, and limitations. |
These projects have different levels of maturity. Research experiments, prototypes, demonstrations, and implemented features are documented as such; they should not be read as claims of production readiness.
| Project | Public site | Source |
|---|---|---|
| Sudomarc portfolio | Open site | myportfolio |
| Interact Club Conakry | Open site | Interact-Conakry |
| CrazyCook — restaurant template demo | Open demo | CrazyCook |
| Carbone Notes — editorial blog | Read the blog | carbone-blog |
| BamanaShop — storefront demo; purchases are not enabled | Open demo | BamanaShop |
| Nova Web — agency website codebase | — | nova-web |
The storefront and restaurant links are demonstrations, not live purchase services.
| Area | Practical focus |
|---|---|
| AI and machine learning | LLM application architecture, model internals, PyTorch experimentation, inference workflows |
| Systems and networking | Linux tooling, network discovery, diagnostics, traffic and firewall administration on authorized systems |
| Software products | Web interfaces, static websites, mobile prototypes, application architecture |
| Automation and developer tooling | Python, GitHub Actions, CLI tools, repository workflows, testing |
| Engineering practice | Technical documentation, scoped changes, security-conscious design, reproducible verification |
- Evidence over claims. Separate implemented behavior from prototypes, experiments, plans, and unverified assumptions.
- Understand before building. Clarify requirements, constraints, architecture, and failure modes before implementation.
- Security by design. Treat inputs and external output as untrusted; keep permissions and secrets scoped.
- Verify the result. Use relevant tests, runtime checks, and reviewable evidence instead of assuming that code works.
- Keep systems understandable. Prefer explicit interfaces, focused changes, and documentation that records limitations.
Projects, experiments, and engineering notes: GitHub.




