Gil Levy
Product · Technology · AI

From an idea to a live product, end to end.

I'm Gil Levy. I built and launched Qualtix, a FinTech SaaS product with over 100 registered users and its first paying customers. I led the product decisions, requirements, AI-assisted implementation, testing and releases, taking it from concept to real users.

See the work LinkedIn Qualtix
qualtix.app
A Qualtix stock analysis: one rating, with separate scores for business quality, entry timing and valuation
A Qualtix analysis. One rating, backed by separate scores a user can read in seconds.
100+ registered usersAnd its first paying customers
About 1,000 US stocksScored automatically every night
Built independentlyFrom product definition to launch
AI-assistedI own the decisions. AI speeds up the build
Featured project

Qualtix

A stock research product for long-term investors. Type a company, and in seconds it answers three questions: is this a strong business, is the price reasonable, and is now a sensible time to buy.

The problem
Individual investors spend hours across reports, ratios and opinions and still end up unsure. Most tools either show raw data or give an answer with no reasoning behind it.
What I built
A rules-based model that scores each company, a page that explains the result in plain language, a daily screener, portfolio and watchlist tools, alerts and email reports, subscriptions with a free trial, and the admin tools to run it all.
My role
Everything a product needs from one person: defining the problem and the requirements, choosing what to build first and what to leave out, directing the build, testing, releasing, supporting users, and deciding what to change next.
Built with
React, Node.js, PostgreSQL, REST and financial-data APIs, scheduled jobs, email automation, and OpenAI for explanations. Development with Claude Code and Codex.
qualtix.app / screener
The Qualtix screener: a list of US stocks with scores, valuation and rating
The screener. Stocks are re-scored each night with the same rules.
qualtix.app
The Qualtix home page with a single search box
The home page. One action: enter a company.
Feb 2026Idea and first working version
Mar 2026Moved to my own server and database. Nightly screener running
May 2026Early access opens. Subscriptions go live
Jun 2026Free trial, onboarding and funnel tracking
Sep 2026Redesigned analysis page, with a fair-value range
Selected work

Building, designing, launching

Three case studies. Each covers the opportunity, the decision I made, what changed, and the outcome.

01

From idea to launch: building Qualtix

How a personal frustration became a live product: defining the core, the decision to keep AI out of the rating, choosing what to leave out, and launching in stages.

Product definitionPrioritisationLaunch
Read case study →
02

Designing for the everyday investor

Users found the analysis page hard to read. I rebuilt it so the answer and its reason come first, in plain language, with the detail in the order a person would ask for it.

UXInformation hierarchyUser feedback
Read case study →
03

Designing the first user experience

What should a new visitor be able to do before the product asks for anything? I designed the path from landing to first value, decided when to ask for sign-up, and built the measurement for it.

ActivationOnboardingProduct analytics
Read case study →
How I work with AI

Product ownership. AI-assisted execution.

Qualtix was built with AI coding agents, mainly Claude Code and Codex. I own the product: what gets built, how it should behave, and whether it is good enough to ship. The agents speed up the execution.

What I do

  • Define the problem and write the requirements
  • Choose the approach, set the technical direction and the priorities
  • Review the result, test the edge cases and investigate what looks wrong
  • Approve every release, and own what happens after it

Where AI agents help

  • Investigate the code and the data, and report what they find
  • Propose options with their risks
  • Write and change the code I approve
  • Run the checks and document the result

How a change ships

1. A written requestWhat should change and why, with the evidence and the risks.
2. InvestigateAgents read and report. At this stage they change nothing.
3. BuildStarts only after my written approval for that specific task.
4. CheckThe build, automated checks, and a comparison with the live version where behaviour could change.
5. Release and recordA separate approval from me. The change and its evidence are written down.

The model decides. AI only explains. The same thinking shaped the product. I wanted the rating to be deterministic and open to inspection: the same inputs give the same result, and a rating can be traced to rules that can be tested. So the rating comes from a rules-based model, and AI is used only to explain it in clear language.

Capabilities

What I do well

Product thinking

Turning a vague problem into a clear requirement, deciding what goes first, and saying no to what does not serve the goal.

Delivery

Taking work from plan to production: scope, QA, release, a way back if something goes wrong, and follow-up afterwards.

Data and analysis

SQL and PostgreSQL, Python and Pandas, dashboards, and the habit of checking whether a number can be trusted before acting on it.

AI and automation

Directing AI coding agents, writing prompt and instruction logic, a user-facing AI assistant, scheduled jobs and email automation.

Systems and APIs

REST and financial-data APIs, authentication, subscriptions and payments, and troubleshooting issues in production.

Operations and growth

Onboarding, pricing and trial flows, funnel tracking, SEO, small paid-search experiments, and working with creators and publications.

Contact

I build products where finance, technology and AI meet, and turn complex problems into experiences people can use.

Email me LinkedIn