The EB1A is a US employment-based first-preference immigrant visa (green card) category reserved for people with extraordinary ability in sciences, business, education, arts, or athletics. Its biggest advantage is that you can self-petition — no employer sponsor, job offer, or labour certification (PERM) is required. You apply by filing Form I-140 with evidence of sustained national or international acclaim. For Indian tech professionals, it is often the preferred green card category because wait times are usually shorter than in EB2 and EB3.

You must show extraordinary ability either through a one-time major achievement, such as an internationally recognised award, or by meeting at least 3 of 10 criteria defined by USCIS. The ones most commonly used by tech professionals include awards or prizes for your work, authorship of scholarly articles, judging or reviewing the work of others, original contributions of major significance, published media coverage about you, memberships requiring outstanding achievement, high remuneration compared to peers, and a leading or critical role in a distinguished organisation. The evidence must be documented independently, not just claimed.

The typical path is: assess your profile against the criteria, spend time strengthening weak areas (publications, citations, judging invitations, talks, press coverage), then build the petition with strong recommendation letters and file Form I-140, which you can self-petition. After approval, you either file I-485 adjustment of status if you are in the US or go through consular processing from India, while tracking the visa bulletin for your priority date. Most successful applicants start preparing 12–18 months before filing so the evidence has time to build.

In two stages. First, USCIS checks whether you meet at least 3 of the 10 criteria (or have a major one-time achievement). Second, in the final merits determination, the officer decides whether the total evidence places you among the small percentage at the very top of your field. Meeting 3 criteria on paper alone is not enough — the quality and independence of evidence matter. A practical self-check is to count verifiable, third-party proof such as citations, media mentions, and judging invitations; if you fall short, a realistic evaluation, ideally from someone who has received an EB1A themselves, helps you identify gaps before you file.

Yes. Because of the per-country limit, India-born applicants face waiting periods for the final green card stage even after the I-140 is approved. EB1 has historically moved faster for Indians than EB2 or EB3, but retrogression does happen, so check the monthly visa bulletin for the India EB1 cut-off date. Filing your I-140 early is important because it locks your priority date, which determines when you can complete the final step.

They serve different goals. EB1A is an immigrant visa — a green card — that you can self-petition for without an employer. O1 is a temporary non-immigrant work visa that needs a US employer or agent to sponsor you, granted initially for up to three years. The O1 bar is generally considered somewhat lower, so a common strategy is to enter or stay in the US on an O1, build stronger evidence while working, and then file the EB1A. If your profile already shows sustained acclaim, going directly for the EB1A saves time; if not, the O1-to-EB1A bridge is the more realistic route.

It broadly has three components: government fees for Form I-140 (with optional premium processing for a faster decision), the green card stage fee (adjustment of status or consular processing), and professional fees, since attorney-drafted EB1A petitions typically cost anywhere from a few thousand to over ten thousand dollars depending on complexity. Add evidence-building costs such as credential evaluations and translations. A professionally handled case commonly lands somewhere between roughly $8,000 and $20,000, and exact figures change periodically, so verify the current USCIS fee schedule before budgeting.

Learn a repeatable framework instead of memorising solutions: clarify the business problem and metrics, define the data and feature pipelines, choose and train models, set up offline and online evaluation, then cover serving, latency, scalability, monitoring, and retraining loops. Practise standard case studies such as building a recommendation system, search ranking, news feed ranking, ads click-through prediction, and fraud detection. Since this round is typically asked at mid and senior levels at large tech companies, doing mock interviews where you defend trade-offs out loud is one of the highest-value parts of the preparation.

"Machine Learning System Design Interview" by Ali Aminian and Alex Xu is the most popular choice because it is written specifically around interview case studies and walks through end-to-end solutions. Pair it with "Designing Machine Learning Systems" by Chip Huyen if you want deeper production and infrastructure thinking, since senior interviewers often probe beyond the standard template. Reading alone is not enough — after each chapter, attempt to sketch the design yourself and practise explaining it in 35–45 minutes, which mirrors the actual interview format.

Cover the five areas almost every process tests: SQL and Python coding, probability and statistics (including hypothesis testing and A/B testing), machine learning fundamentals such as the bias-variance trade-off and evaluation metrics, a product or case-study round, and behavioural questions. A focused 6–8 week plan works well — daily SQL and statistics practice, two to three full mock case studies per week, and prepared STAR stories for your projects. Tailor the depth to the company type: product companies weight metrics and case thinking heavily, while analytics-heavy roles weight SQL and statistics.

They usually fall into a few buckets: SQL questions on joins, group-bys and window functions; probability puzzles around dice, cards and conditional probability; statistics questions on p-values, confidence intervals and how to design an A/B test; machine learning questions on overfitting, regularisation, handling imbalanced data and choosing the right metric; and open-ended cases like "how would you measure the success of a new feature" or "a key metric dropped 10% — how do you diagnose it?" Practising your reasoning out loud matters as much as getting the answer right.

As a fresher, expect fundamentals rather than deep specialisation: basic statistics and probability, Python with pandas, SQL, classical algorithms like linear and logistic regression, decision trees and clustering, and knowing which metric to use to evaluate a model. Interviewers pay close attention to your projects, so be ready to explain the problem, your choices and the impact of anything on your resume end to end. Internship work, portfolio projects and clearly explained trade-offs usually matter more than knowing the latest deep learning architectures.