Research & Projects

SetupX overview from the paper
RESEARCH SYSTEM

SetupX ↗

From a repository URL to a verified environment. Experience retrieval, speculative execution, and adversarial verification.

LLM agentsDockerXPU
Workspace-Bench 1.0 overview from the paper
COLLABORATIVE RESEARCH

Workspace-Bench 1.0 ↗

A benchmark for agents working across realistic workspaces and large-scale file dependencies.

EvaluationWorkspace learning
LLM/Agent-as-Data-Analyst overview from the paper
SURVEY & RESEARCH RESOURCES

LLM/Agent-as-Data-Analyst ↗

Research across data modalities, including my contributions on video and 3D-model analysis.

Multimodal AIData analysis
COURSE PROJECT

SpeechMaster ↗

Budget-aware self-supervised speech recognition with complementary model routing and representation-compression experiments.

Speech recognitionPyTorch
COURSE PROJECT

KV Cache Compression ↗

Comparing StreamingLLM, SnapKV, and PyramidKV on Pythia-70M: quality, memory, and latency trade-offs.

LLM inferenceEvaluation
SCIENTIFIC TOOL

MS/MS Annotation Platform ↗

A local application for blind and model-assisted structure annotation, peak-level evidence, and exportable results.

Scientific AIAnnotation