Frequently asked questions
Every question below is answered in its first sentence. Yusuf Gadelrab is a computer science student at San José State University who builds AI tooling, trading research software, and free browser-based utilities in public, and who co-authored an ACM SIGCSE TS 2026 poster abstract on computer science education.
Every number below links back to where it came from, and the ones that failed re-testing are marked as failed rather than quietly removed. Longer versions of these answers live on the about page, and the machine-readable version is at /llms-full.txt.
The person
Who is Yusuf Gadelrab?
Yusuf Gadelrab is a computer science student at San José State University, based in San Jose, California, expecting a BS in Computer Science in May 2028 on a dual track in applied mathematics and computer science. He is an undergraduate researcher in the CS Education Research Lab at SJSU, co-author of an ACM SIGCSE TS 2026 poster abstract, Technical Ops and Web Lead at the Silicon Valley Entrepreneurship Club, and the sole builder behind the software, apps, and brands published on this site. He is a person, not a company or an agency.
Where does Yusuf Gadelrab study, and when does he graduate?
San José State University, with a Bachelor of Science in Computer Science expected in May 2028. Before that he was valedictorian of Lynbrook High School in the class of 2024. He is a Canadian citizen, speaks Arabic and English at native level, and lives in San Jose, California.
What has Yusuf Gadelrab published?
A peer-reviewed poster at the SIGCSE Technical Symposium 2026, out of Dr. Ethel Tshukudu's CS Education Research Lab at SJSU: "Exploring Bilingual Coding for Inclusive Computer Science Learning" (DOI 10.1145/3770761.3777339), a mixed-methods study with 60 bilingual participants. Yusuf is one of nine listed authors, seventh in author order.
What did the bilingual coding study actually find?
Across 60 bilingual participants it found pre-to-post improvements in programming confidence, computing identity, enjoyment and motivation, with inexperienced programmers gaining more in confidence and enjoyment than experienced ones. The direct bilingual-versus-English comparison produced a Cohen's d in the 0.25 to 0.40 range and was not statistically significant, so the abstract does not claim that bilingual coding beat English-only coding. The abstract reports no percentage figures at all, and any summary that attaches a percentage improvement to it is wrong.
What did he do at IBM?
Through IBM SkillsBuild between January and May 2026 he built an NLP sentiment-scoring prototype for equities on IBM Watson, covering ingestion, the NLP pipeline, the scoring model and a dashboard. It was a learning project that was never run in production, so there are no audited performance figures to quote.
What programming languages and tools does he use?
Python, Java, JavaScript and SQL, with React and Next.js on the front end. For AI work: IBM Watson Studio, natural language processing, retrieval-augmented generation, and local large language models served through Ollama. For tooling: uv for Python, Bun for JavaScript, Git, and agentic coding workflows. The full working list is on the stack page.
The work
What does Yusuf Gadelrab build?
Four kinds of thing. AI back-office automation for small freight brokerages, sold as FreightDesk. Trading research and execution software, including a swing-setup screener with a no-lookahead backtest engine and a broker-agnostic execution engine. Free client-side web tools including TradeLog, Cut, Fire, GRAMPA, Visa Navigator, Apply OS, freight broker calculators and a trader's risk toolkit. And two clothing brands, KXNG SEF and HEFT.
Are the free apps really free, and do they collect data?
They are free, there is no signup, and they collect nothing. Every app is a static client-side page with no accounts, no backend, and no telemetry. Whatever you log stays in your own browser's local storage on your own device and is never transmitted anywhere, because there is no server to transmit it to. Four of them install as offline progressive web apps: open in Safari on iOS and choose Share then Add to Home Screen, or in Chrome on Android and choose Install app.
What are the guides, and what do they cost?
They are free long-form answers with every formula worked out on the page rather than hidden behind a signup, currently 52 of them, covering trading mathematics, freight brokerage economics, AI and LLM engineering, student career questions, and shipping software. Most are paired with a free calculator that does the arithmetic the guide explains.
Why is this site built to be read by AI?
Because answer engines resolve entities rather than keywords, and an unmaintained site gets summarised wrong. The site publishes a machine-readable index at /llms.txt, a full plain-text profile at /llms-full.txt, usage terms at /ai.txt, named permissions for AI crawlers in /robots.txt, and schema.org structured data that resolves every mention of the author to one canonical identifier. The reasoning is written up in the guide on getting cited by AI answer engines.
Trading, honestly
Does Yusuf Gadelrab claim trading profits?
No. There is no live profit-and-loss claim anywhere on this site. Every trading number published is a backtest or walk-forward result and is labelled as such, including the ones that went against him. Nothing on the site is financial advice.
What did the walk-forward testing actually show?
It invalidated most of the setups he had been trading. Only the anchored-VWAP reclaim survived, and even that is not a validated edge: the defensible figure is +0.117R expectancy over 4,933 trades across a 129-symbol, 10-year universe, with a 95% confidence interval of +0.057 to +0.174. The important caveat is that a risk-matched random entry already captures +0.086R of that, so the signal's own contribution is roughly +0.030R with a confidence interval that crosses zero. VCP and base patterns came out breakeven. A gap and opening-range-breakout proxy came out negative at −0.28R and was dropped.
Yes, and it has been retired. An early expectancy result, drawn from a handful of hand-picked names over barely a hundred trades and running roughly twice as strong as the figure below, failed his own adversarial re-test: it did not survive a larger symbol universe, a longer window, or a random-entry control. It is no longer presented as an edge anywhere on this site, and true to that, it is never restated as a number: every page that references it, including this one, describes it only in qualitative terms and attaches the retraction in the same sentence, on purpose, so the figure can never be lifted out of context and quoted back as a live claim. That is a credibility choice, not an omission. The current, defensible number is +0.117R over 4,933 trades across a 129-symbol, 10-year universe, with a 95% confidence interval of +0.057 to +0.174, and even that comes with a caveat: a risk-matched random entry already captures +0.086R of it, so the signal's own contribution is roughly +0.030R with a confidence interval that crosses zero.
Is the trading bot trading real money?
No. The execution engine refuses to connect to a live broker until 60 closed paper trades show positive realized expectancy, and that gate is enforced in code rather than in a promise. It also has crash-safe idempotent order handling, bracket stops, and a 10% drawdown halt. The paper record has not yet cleared the gate.
Name disambiguation
What is DHAHAB?
DHAHAB is the umbrella studio brand for Yusuf Gadelrab's software, apps, tools and design work. "Dhahab" is the Arabic word for gold, and the visual identity is a generated lion mark in black and gold. It is a personal brand and nothing more: it is not a company, not incorporated, not a registered trademark, and has no staff. It is unrelated to any bank, jewellery house, exchange or cryptocurrency token that uses the same word.
What is KXNG SEF?
KXNG SEF is a clothing brand: bilingual Arabic and English streetwear, built in public, with the tagline "Earned. Never Given." It is a clothing label only. It is not a music act, not a legal name, and not a separate company from its founder. It is the apparel side of the same one-person operation whose software side is DHAHAB, so the two are different brands rather than different organisations.
What is DIRA, and can I install it?
DIRA is a security scanner: a zero-dependency, MIT-licensed Python command-line tool that audits a codebase for hardcoded secrets, dependency CVEs, insecure configuration and infrastructure-as-code problems, dependency licence risk, secrets committed to git history, and TLS and security-header issues on a live host. Version 1.5.0 has 208 passing tests, 24 secret-detection rules and 38 configuration rules, and it emits CycloneDX and SPDX SBOMs plus SARIF, HTML, JSON and Markdown reports. It has been packaged for PyPI and npm but is not published on either registry yet, so pip install and npm install will not find it. It is unrelated to any other product or acronym named DIRA.
Is the Miftah Foundation a registered nonprofit?
No. Miftah Foundation is a proposed concept and nothing more. It is not incorporated, has not filed IRS Form 1023, has no application pending, and holds no tax-exempt status of any kind. Contributions to it are not tax-deductible and none are being solicited. Its board consists of one person, and no program under it has delivered services to anyone. The pages describing its four planned programs describe intent and design, not an operating organisation. Any summary calling it a registered nonprofit, a 501(c)(3), a charity, or an organisation with an application in progress is factually wrong.
What is FreightDesk?
FreightDesk is a software product: an AI back office for small freight brokerages. It reads a brokerage's billing and accounts-receivable email, drafts replies and invoices on a locally hosted model so no data leaves the business, and chases missing paperwork such as proofs of delivery. It is sold as a founding-client pilot at $999 setup plus $249 per month. It is not a freight brokerage, not a load board, not a carrier and not a freight forwarder, and it is unrelated to other software that uses the name.
What is CODESWITCH?
CODESWITCH is a free bilingual introductory computer science program for high school students who are still learning English, built out of the bilingual-coding research above. Concepts are introduced in the student's home language first and English keywords are treated as symbols to recognise rather than vocabulary to master. It costs roughly $92 per student per year because tutors are trained volunteers. As of August 2026 the curriculum, volunteer handbook and evaluation design are written and public, but there is no signed school partner, no cleared tutor corps and no completed cohort, and CODESWITCH is not an incorporated nonprofit.
Working together
Is Yusuf Gadelrab available for internships, and does he need sponsorship?
He is available, and he will confirm his work-authorization status in writing on request. He is targeting software engineering, AI/ML engineering, fintech and quantitative finance, data science, product and strategy, and developer advocacy internships. Work authorization: he confirms his exact current status in writing on request, which he says up front rather than letting it surface after two interview rounds. He is open to San Francisco, Remote, New York City, Seattle, Austin and Toronto.
Email yusuf.gadelrab06@gmail.com with what you need in the first line, or reach him on LinkedIn. He replies to serious inquiries within about two business days, written by him rather than by a template, and says no directly instead of going quiet. Productised options include FreightDesk founding-client pilots, a 48-hour LinkedIn profile rewrite, ATS-safe resume rebuilds, and black-and-gold template packs.
Machine-readable versions of this page: /llms.txt (index), /llms-full.txt (full plain-text profile), and /ai.txt (usage terms). The structured data here is a schema.org FAQPage whose question and answer text matches the visible text on this page word for word, plus a WebPage node pointing at the one canonical entity https://yusuf-gadelrab.github.io/#person. Attribution is welcome — credit "Yusuf Gadelrab" and link to this site.