Case studies
Long-form technical records of production systems and field work. They stay close to implementation detail across architecture, frontend and backend, data, infrastructure, difficult edge cases, and AI when it was part of the build.
01
Full-Stack Product / Internal Workspace / AI Execution Layer
Internal Sales Operations Workspace
A custom full-stack workspace that connects lead research, CRM context, offers, follow-ups, deadlines, documents, client briefs, delivery notes, and playbooks into one execution layer for sales and delivery teams.
02
Document Processing / OCR / Vision-Language Models
Self-Hosted Multi-Model OCR/VL Document Preprocessing Pipeline
A self-hosted OCR/VL preprocessing pipeline for turning messy technical documents into structured, retrieval-ready data for downstream AI systems.
03
Data Engineering / Web Extraction / AI-Assisted Structuring
Large-Scale Web Data Extraction Pipeline
A large-scale web data extraction pipeline that filtered a large candidate link set, processed valid targets, used discovery fallbacks, and produced structured data for downstream systems.
04
Retrieval Systems / Hierarchical Documents / Source-Grounded Answers
Hybrid RAG System for Hierarchical Documents
A retrieval system for complex structured documents that preserved hierarchy, combined BM25 and vector search, used RRF ranking, and returned source-grounded answers.
05
Analytics / Text2SQL / Hybrid Retrieval
AI Analytics Chatbot Combining SQL Aggregations and Vector Retrieval
An AI analytics chatbot that combined validated SQL aggregations from structured database tables with BM25/vector retrieval and reranked context from unstructured knowledge sources.