Portfolio
Systems, research, and things that work.
A portfolio of difficult problems made inspectable—from agent architecture and cybersecurity research to the infrastructure beneath them.
Each case study follows the problem, constraints, approach, architecture, outcomes, and next questions.
4 projects
| Project | Kind | Status | Role | Technologies |
|---|---|---|---|---|
| C2 — A Single-Human Company OS | software | building | Architect, builder, and sole operator | Slack front door and human approval, Alli orchestration and Missions, Domain teams: managers, specialists, validators, Per-team MCP tool surface (read and propose), Approvals service with its own identity, C2 Spine: PostgreSQL system of record, governed write API, n8n workflows, C2 Brain: pgvector semantic memory (infrastructure), Containers, reverse proxy, tunnel, private media archive |
| Transformer-augmented reinforcement learning in the Unity Food Collector environment | experiment | complete | Author of the PPO transformer-actor variant and the PPO baseline; architecture, training loop, stability fixes, experiment configuration, and report | Unity ML-Agents Food Collector (5×40×40 grid sensor, hybrid action space), Gymnasium wrapper over the Unity binary, PyTorch, Proximal Policy Optimization with GAE and clipped surrogate, CoordConv-ResNet visual encoder (31.1M parameters), Transformer encoder actor (d_model 512, 2 layers, 2 heads), Decoupled MLP critic on the raw observation, Cosine learning-rate schedule with 10k-step warmup, Aim experiment tracking |
| Claude Kit: domain templates that turn Claude Code into a specialist | tooling | archived | Designer and builder | Claude Code, Node.js scaffolding and validation scripts, Shell hooks, Markdown agent, skill, and rule definitions, MCP servers: context7, GitHub, arXiv, Brave Search, Neo4j, pgvector |
| Deepfake detection across generations: MesoNet to ViT | research | complete | Co-author; MesoInception4 and CBAM models, the 17-run ablation, cross-dataset evaluation, and diffusion dataset assembly | PyTorch, MesoNet / MesoInception4, CBAM attention with Grad-CAM, Xception and ViT-B/16 fine-tuning, LVNet two-stream model, MTCNN face cropping, FaceForensics++, Celeb-DF v2, DiFF, DeepFakeFace |