active studies
ongoing research and engineering studies — benchmarking, evaluating, and building at the frontier of ai systems.
openclaw enterprise evaluation
evaluating nvidia's nemoclaw and openshell for enterprise agentic deployment. studying privacy router architecture, local-first inference with nemotron models, and sandbox security patterns.
1-bit inference benchmarks
benchmarking microsoft's bitnet b1.58 models across arm and x86 hardware. profiling energy consumption and comparing against fp16/bf16 baselines for edge deployment scenarios.
structured compression with openzl
evaluating meta's openzl format-aware compression framework for ai workloads. testing compression ratios and throughput on model weights, training data pipelines, and structured datasets.
zero-knowledge credential systems
studying google's longfellow zk library for privacy-preserving identity verification. ecdsa-based zero-knowledge proofs, eu eidas compliance, and potential applications in agentic authentication.