Built your app with Antigravity, Cursor, Lovable or Bolt, or running a database that reports slowly or "too clean"? I review the code, test every key flow and give you a plain-English fix list. 11+ years building and testing enterprise software.
Taking projects now: a few audit slots and one monthly QA retainer.
status=x' OR '1'='1 on a test database. Every order comes back.This is how every finding in my reports reads: where, how to reproduce, why it matters, how to fix. Example from a sample report.
Fixed scopes and written quotes. Tell me what you need and I will send a price before any work starts.
Code review, full manual test pass, security basics and a prioritized report with screenshots and steps to reproduce. A walkthrough call and one retest are included.
Slow queries and stored procedures, indexing, data cleanup, migrations and reconciliation checks, with measured before and after results.
Test plans, regression and exploratory testing, bug triage and release sign-off, as a monthly retainer or per release.
One practical AI feature or workflow added to your app, with fallbacks, logging, cost limits and tests.
Templates, inventories and config-driven deployments for application teams, so releases are repeatable and manual errors drop.
Personal projects I designed and built myself.
An engine that uses Google Gemini to write product campaigns and publishes them through the Meta Graph API. I built a prompt-packaging and context-parsing workflow to keep token use down, and checked the REST and SOAP feeds with SoapUI.
I verified output on web and mobile for layout, attribution tracking and link behavior.
It includes a short-link service I built on Cloudflare Workers, which creates social link previews and sends readers on to the destination page.
See it running on its public Facebook page (posts contain affiliate links).
Python, Gemini API, Meta Graph API, Cloudflare Workers, REST, SoapUI
An event-driven Python application that handles live price ticks and order dispatch with asyncio and WebSockets. It has multi-broker authentication (TOTP and OAuth2), risk controls, an SQLite telemetry store and a desktop analytics screen.
Python 3.12, asyncio, PyQt6, SQLite. Engineering showcase only; not investment advice or a product for sale.
Both pieces below are built so you can see how I think. Neither is a client project.
On a 5-million-row ledger, one NOT IN and a few NULLs hid every ledger-only record. The same job also scanned both tables because of a date function and a text-type mismatch.
I show the fix, the scripts, and a row-by-row check that proves the new query matches the old one where it should.
Ask for the case studyFifteen findings from critical to low: customers reading each other's orders, prices accepted from the browser, exposed secrets and duplicate payments. Each has a fix and a ready prompt for your AI tool.
It shows what you receive: summary, findings, action plan and retest terms.
Ask for the sample reportFour steps, all in writing.
I'm Sahil. For 11+ years I have built, led and tested enterprise software, with a focus on data accuracy, SQL performance and release quality. I led a development team through Agile releases and received several Star Performer awards. My M.Tech thesis and a 2015 conference paper were on information security.
Today I work with founders and small teams. I use AI development tools myself, so I know where they go wrong, and I test the output the way a QA lead would.
A review like this is a quality check with security basics. It is not a penetration test or a security certification, and I say so in every report.
I'll reply with how I would approach it and, where possible, the first issues I would check.