The Pipeline at a Glance
All ten stages from resume screen to offer, typical timelines, where mid-level candidates most often fail, and how to run parallel processes.
The whole pipeline in one view
A mid-level (L4 / E4 / SDE2 — different companies' labels for the same "can own a feature end-to-end" level) interview process at a large tech company runs through up to ten stages. Most candidates only prepare for two of them — the coding rounds — and then get surprised by the other eight. Here is the full map:
| # | Stage | Typical duration |
|---|---|---|
| 1 | Application / resume screen | Days to 2 weeks for a response |
| 2 | Recruiter screen | 20–30 min call, scheduled within 1–2 weeks |
| 3 | Online assessment (OA) | 60–90 min, within a 3–7 day window |
| 4 | Technical phone screen | 45–60 min live coding |
| 5 | Take-home / practical project (at some companies) | 2–8 hours of work over 2–7 days |
| 6 | Onsite loop | 1 day of 4–6 rounds, 1–3 weeks after the phone screen |
| 7 | AI-assisted coding round (new, at some companies) | 60 min, inside the onsite loop |
| 8 | Debrief / hiring committee | 1–2 weeks after the onsite |
| 9 | Team match (mostly Google, sometimes Meta) | 2–4 weeks, can stretch to months |
| 10 | Offer and negotiation | 1–2 weeks from verbal to signed |
End to end, application to signed offer commonly takes 8–12 weeks. Not every company runs every stage — Amazon usually hires you straight into a team (no team match), and many companies skip the OA for experienced candidates — but you should assume the long version and be pleasantly surprised.
Stage 1 is mostly machines now
Roughly 78–80% of large tech firms run AI or ATS pre-screening (ATS = applicant tracking system, the software that files and filters resumes) before a human reads anything. Modern screeners are smarter than old keyword matchers: listing "machine learning" as a skill with no work history that shows it gets discounted, not counted.
The single biggest lever at this stage is not resume polish — it is a referral. Referred candidates are about 4x more likely to get an offer and are hired roughly 55% faster than cold applicants, largely because a referral skips the machine filter. A good rule for your time: spend about 60% of job-search hours on networking and direct outreach, 40% on applications.
Where mid-level candidates actually fail
Knowing the failure points tells you where to spend prep time:
- Level calibration at the recruiter screen. The recruiter decides whether to run you as L4 or L3 partly from how you frame your experience. Say "I was on the team that built X" and you sound junior; say "I designed and shipped X end-to-end" and you sound mid-level. This 30-minute call sets your ceiling before any technical round happens.
- The OA under proctoring. Assessments in 2025–2026 are heavily monitored (details in the next lesson). A flagged session ends a candidacy — nothing you do later recovers it.
- Follow-up questions in live rounds. Because AI cheating is widespread, interviewers now drill into "why" — why this data structure, modify your solution for this variant. Candidates who pattern-matched a memorized answer fail here even after producing working code.
- Down-leveling at debrief. A borderline packet at L4 often produces an L3 offer instead of a rejection. This is common, and coding performance is weighted most heavily in the L3-vs-L4 split — which is why the mid-level coding bar (covered in its own module) matters so much.
- Team-match expiry. At Google, passing the hiring committee puts you in a pool, and your packet expires after about 8 weeks if no team picks you. People genuinely pass everything and still get no offer in slow quarters.
Run processes in parallel — deliberately
The strongest negotiation lever that exists is a competing offer. That only works if offers land at roughly the same time, so treat scheduling as part of your strategy:
- Batch your applications. Apply to your full target list within the same 1–2 weeks, not one company at a time.
- Warm up first. Put one or two companies you care less about early in the schedule. Your first onsite is rarely your best.
- Control pacing with recruiters. It is completely normal to say "I'm mid-process with a few companies and expect decisions around the week of the 15th — can we schedule the onsite before then?" Recruiters do this dance daily; they will compress or stretch for a candidate they want.
- Keep everything warm through team match. If one company parks you in a matching pool for a month, active processes elsewhere are your insurance — and your leverage.
Aim for all offers inside the same 2-week window. That is the setup that makes the negotiation stage (stage 10) actually work.
How to use the rest of this module
The next three lessons go deep on the stages where preparation changes outcomes: the early screens and assessments, the onsite loop and what happens after it, and the new AI-assisted rounds that Meta and Google introduced in 2025–2026. The behavioral and mid-level-bar modules then cover what you say in those rooms. Keep the ten-stage map above in mind — every later lesson slots into it.