Index
Selected work, with the evidence attached.
Three case studies, written the way I’d want to read them: the problem, the constraints, what I did, how it was measured, and what the numbers don’t show.
01 / Case studies
Case studies
- ARetrieval evaluation · researchLocalRAG++An offline, CPU-only retrieval pipeline and evaluation harness: sparse, dense and hybrid retrieval, cross-encoder reranking, a latency sweep, and a separate study of the document quality that comes before retrieval.
- BExtraction · validation · industry workFinancial document intelligencePulling specific figures out of long annual reports and making sure they agree with a reference before anyone relies on them: PDF-first extraction, table-aware chunking, schema checks and reconciliation.
- CSaaS backend · collaborative buildColeroon.aiCo-developing a SaaS system where most of the hard parts are other people’s APIs: OAuth integrations, background jobs, quota and cost tracking, publishing workflows and the observability to see them.
02 / Smaller & earlier
Without write-ups
Listed for context. These get a page when there’s public material to show.
- GPT-2-oriented components, rebuiltBluBridge TechnologiesExploring Rust/CUDA and custom PyTorch numerical layers during a research internship.
- Real-time vehicle detectionIntel UnnatiYOLO-based detection.
- Data pipelines on Azure DatabricksInternship workPipeline and API development.
- AI-assisted study toolGoogle × GeeksforGeeks hackathon5th of 608 teams.