Problem: Clinical notes contain high-value details that are difficult to reuse for research databases or review workflows.

Built: Built an MVP for extracting structured clinical information from notes or mock records with JSON validation and human-review-friendly output.

Impact: Demonstrates a practical clinical informatics path from narrative text to auditable structured data.

Problem: Homopolymer-region artifacts and flow-rate or fluidics issues can complicate variant review.

Built: Designed a dashboard framework for automated amplicon coverage monitoring, sequencing artifacts tracking, and clinical sequencing QC investigation.

Impact: Supports faster investigation of false-positive results and run-level sequencing behavior by pathologists on a no-code interface.

Problem: Variant calls from different platforms and callers rarely align cleanly for downstream clonal evolution tools.

Built: Planned a harmonization workflow for caller-specific parsing, normalization, and ML-ready tabular transformation.

Impact: Improves compatibility across sequencing workflows and downstream clonal tracking analysis.

Problem: PhIP-Seq peptide-level reactivity needs reproducible modeling to support cancer research interpretation.

Built: Created a statistical scoring framework for peptide and antigen-level reactivity in a colorectal cancer research context.

Impact: Makes antibody profiling outputs easier to compare, summarize, and investigate.

Problem: Variant review requires comparing evidence across multiple clinical and public knowledge sources.

Built: Designed an AI-assisted review pattern with ClinVar/dbSNP/COSMIC-style placeholders, evidence comparison, and human-in-the-loop summaries.

Impact: Frames a practical assistant for clinical reporting support without replacing expert review.