RESEARCH MONOGRAPHS
In-depth technical papers on state graph recursion, token efficiency, small model distillation, and deterministic agent harnesses.
Zero corporate fluff or hand-waving. Each monograph breaks down specific engineering bottlenecks, mathematical invariants, synthetic datasets, and production system implementations.
PEER-LEVEL ESSAYS
An architectural dissection of why DAGs and linear pipeline chains break under real-world tool errors. Covers cyclic recursion limits, checkpointing, state reducers, and how to engineer deterministic self-correction loops.
How we compressed operating system automation into SmolLM2-360M and Qwen2.5-0.5B. Explores synthetic multi-turn action trace curation, 4 mechanical verification gates, and eliminating runtime dependency bloat with pure Python standard-library harnesses.
CORE INVESTIGATION DOMAINS
01 // CYCLIC MULTI-AGENT SYSTEMS
State graph topologies, checkpointing persistence, human-in-the-loop interruption mechanisms, and hierarchical supervisor coordination models.
02 // EDGE MODEL DISTILLATION
Transferring frontier reasoning into sub-1B parameter models via synthetic trajectory verification, parameter-efficient fine-tuning (PEFT), and GGUF quantization.
03 // DETERMINISTIC KNOWLEDGE RETRIEVAL
Hybrid search fusing dense vectors with Okapi BM25 sparse indices, Graphify AST graph walks, and Torikago jailkeeper sandboxing.