Tools I Use
Claude (Anthropic)
My primary AI coding partner — used throughout building this portfolio itself, from component architecture to debugging Supabase RLS policies.
ChatGPT / GPT models
Content drafting, brainstorming copy and structure, and reasoning through architecture decisions before implementation.
GitHub Copilot
In-editor code completion for faster day-to-day implementation once a pattern is established.
AI-assisted data processing
Cleaning, structuring, and summarizing data pulled from multiple sources — the same pattern behind KarJo's job aggregation.
My Approach to AI Integration
AI drafts, I decide
AI is fast at generating options — code, copy, structure — but every architectural decision, every line that ships to production, is reviewed and understood by me first. I don't ship code I can't explain.
Automate the repetitive, not the judgment calls
Data collection, formatting, first-draft content, boilerplate code — good candidates for AI. Security decisions, database schema design, and client-facing architecture — I keep those fully in my own hands.
AI as a force multiplier for a small team
As someone who leads a small technical team, AI tooling lets us move faster on execution while keeping the same headcount — more useful in a market like Afghanistan's where hiring specialized talent is harder.
KarJo — Automated Job Aggregation
KarJo automatically collects job listings from multiple sources — job boards and employer submissions — into a single feed, with a scheduled daily digest sent via push notification and Telegram. This is where I'm applying automation today: pulling, cleaning, and structuring data at scale so a user never has to check five different places.
Where I want to take it next: adding an LLM layer on top of the raw aggregation — so instead of just collecting listings, the bot could match a candidate's saved profile against new postings and flag the ones actually worth applying to, or let someone search in plain language ("remote frontend jobs this week") instead of manually filtering. That's the difference between an aggregator and an assistant.
Read the full KarJo case study →
Try It: KarJo Smart Matching Preview
A simulated preview of the AI matching concept described above — these example matches are pre-written, not live results from a model or KarJo's actual database. Click any result to browse real, current listings on KarJo.
AI Draft Assistant for This Portfolio's Own Dashboard
The admin dashboard behind this site already lets me manage blog posts, projects, and case studies. The next step is adding an AI draft button directly inside the blog editor — I'd give it a topic and a few bullet points, and it would generate a first-draft post in my voice, which I'd then edit and publish. It's a small feature, but it's a real, buildable example of AI reducing the time between "I have something to write about" and "it's published."
What's Next
Smarter job matching in KarJo based on a candidate's profile and application history, not just keyword search
Natural-language job search — 'remote frontend jobs posted this week' instead of manual filters
An AI draft assistant inside this portfolio's own admin dashboard, for generating first-draft blog posts from a few bullet points
Automated case-study summarization for future client projects, cutting the write-up time after a project ships