I build and study search, retrieval, ranking, and the systems behind modern AI search.
I’m a developer focused on search and retrieval systems — especially how results are retrieved, ranked, evaluated, and monitored.
A lot of what I build starts with a simple question:
Did the search results actually get better — or did the ranking just change?
That question has taken me deeper into search relevance, vector retrieval, reranking, evaluation, embeddings, and the infrastructure behind modern AI search.
A search-quality, ranking-evaluation, and observability project built around Exa.
I’m using it to explore ranking changes over time, snapshots, result comparison, retrieval quality, and the signals that can help explain when search behavior changes.
A local-first search evaluation workspace built around a specific problem: when a reranker changes the order, did it actually improve relevance?
Exa retrieval → Cohere reranking → blind human judgments → NDCG / RR
The goal is to keep retrieval, reranking, and evaluation separate — and compare rankings against independent human relevance labels instead of treating model scores as ground truth.
I’m also experimenting with Weaviate as a retrieval layer and exploring candidate generation, vector search, hybrid retrieval, reranking, and search evaluation.
- Search relevance and ranking evaluation
- Sparse, dense, and hybrid retrieval
- BM25 and embeddings
- Cosine similarity, KNN / ANN, and HNSW
- Candidate generation and reranking
- NDCG, Recall@K, MRR, and relevance judgments
- Vector databases and retrieval infrastructure
- Search observability, ranking drift, and quality measurement
I care about understanding the systems I build — not just getting an API response back. I want to know why something works, where it fails, and how to measure whether a change actually helped.
TypeScript · Next.js · React · Node.js · Weaviate · Exa · Cohere · Redis · PostgreSQL · Appwrite · Tailwind CSS
I’m spending most of my time getting better at search relevance, retrieval, ranking, vector databases, and AI infrastructure — building experiments, reading how real search systems work, and turning what I learn into projects.
If you're working on interesting search or retrieval problems, I’m always happy to talk.


