Portfolio

Projects and Writings

Projects

Knowledge Navigator

A RAG-powered Q&A tool built on the AI That Works podcast archive. Ask questions, get answers with source citations from the show.

Hybrid RAG
Contextual Retrieval
Agentic Data Pipeline
Python
Next.js
FastAPI
Neon DB
Vercel
GCP

View Project

Job Search Automation

An agent plugin that crawls target company job boards, scores postings against a rubric, and surfaces matched jobs for review in a local dashboard.

Agent Plugin
Agent Skills
Agentic Workflow
MCP
TypeScript
Node.js
SQLite

View Project

Writing

Your AI Breaks on the Way to Production

July 30, 2026

Your AI Breaks on the Way to Production AI works in the demo. It breaks on the way to production. It breaks at three distinct boundaries. You need an AI evals strategy that covers all three. Because the failure modes are different. The inve...

ai-applications
harness-engineering
ai-agent
ai

Read on Medium

Same Prompt, Model Upgrade, Half the Answer

July 16, 2026

Same Prompt, Model Upgrade, Half the Answer Why upgrading your LLM creates regressions that your test suite can’t catch You upgrade from gemini-3-flash-preview to gemini-3.5-flash. Same family. Same API. Same prompt. Both pass your test sui...

generative-ai-tools
llm
ai
genai

Read on Medium

Beyond Basic RAG: Building a Multi-Stage Retrieval Pipeline with Spec Driven Development

May 20, 2026

TL;DR: Single-method retrieval has limited levers to enhance your application. A hybrid retrieval pipeline with cross-encoder gives you the ability to tune precision and recall by casting a wide net first and narrowing aggressively. This ar...

ai-development
genai
ai
generative-ai-development

Read on Medium

How I built MVP Podcast Knowledge Navigator using Spec Driven Development approach

May 7, 2026

TL;DR How I built an MVP knowledge navigator using a Spec-Driven Development approach. The build took roughly 1–1.5 weeks from initial research to local verification, plus another week to handle hosting (i.e. figuring out service accounts i...

genai
product-development
spec-driven-development
ai

Read on Medium

Human+AI: How You Use AI Shapes Whether You Grow or Stagnate

March 9, 2026

TL;DR: How you use AI can help or hinder how you learn new skills. A recent Anthropic study shows a 62 point mastery gap predicated on how you use AI. Users who stayed cognitively engaged learned new skills; those who delegated didn’t. This...

agentic-ai
ai
ai-agent
generative-ai-tools

Read on Medium

Stop Chasing Bugs Only: Build AI Evals That Anticipate Problems

February 5, 2026

TL;DR: Proactive evals catch issues before users do Most teams build evals reactively — wait for failures, scramble to add tests, repeat. This article shows a proactive way: map user journeys, generate test cases with LLM assistance, valida...

generative-ai-tools
ai
genai
llm-evaluation

Read on Medium