About Armel Tenkiang
I’m a Portuguese-American computer scientist and researcher of Fulani and Bamileke origin. I studied at the School of Computer Science (SCS).
I’ve worked in software for more than twenty years, primarily in TypeScript and Python. My interests include local data ownership, recovery after failure, and checking the claims made by models.
Systems practice
With Galidima, my work on local inference covers the provider boundary, failure handling, context budgets, source attribution, and latency and token measurement.
For financial and operational data, I prefer deterministic paths and read-only domain contracts. The model interprets evidence; it does not replace the authoritative record.
Durable state remains in SQLite. Snapshots and backups are written atomically and checked before they are retained elsewhere.
Research Areas
- Zero-knowledge proofs for private state, selective disclosure, and auditability.
- Context shaping for LLM systems: constraints, compression, retrieval boundaries, and bounded memory.
- Local-first sync, CRDTs, and recovery under partial connectivity.
- Verifiable payment systems with private settlement links.
- Append-only coordination logs for multi-agent and human review.
Profiles
GitHub: 249 contributions in the last 12 months, as of ; 157 tracked project commits in 2026.
Listening
Occasional mixes and reference playlists live on SoundCloud.
Selected Technical Notes
Contact
For project notes or technical context, send a short message.
Email