How are others coping with Meta AI layoffs and next steps

General
1

I was recently impacted by the Meta layoffs on the AI side and I’m feeling lost about what to do next. I’d really appreciate advice on navigating the job market, updating my AI portfolio, and targeting companies that are still hiring for similar roles. Any practical tips, resources, or recent experiences from others who went through the Meta layoffs or similar big tech cuts would really help me figure out my path forward.

2

Went through the same thing last year, also from a big-name AI team. Here is what helped, step by step.

  1. Reset your story
    • Write a tight 2–3 sentence “who I am / what I do / what I want” blurb.
    Example: “Applied ML engineer with 4 years on LLM infra and recsys at Meta. Strong in PyTorch, distributed training, and evals. Looking for roles in applied LLMs or ranking at Series B+ startups or larger.”
    • Use that blurb in your headline, summary, and intro emails.

  2. Update your resume for non Meta people
    Meta managers understand internal team names. Recruiters outside do not.
    For each role, do:
    • 1 line: what the team did in plain English.
    • 3–6 bullets: measurable outcomes.
    Examples:
    • “Trained and shipped a ranking model that improved click-through by 3.4% on N users.”
    • “Cut inference cost 18% by quantizing models and improving batching.”
    • “Owned E2E of X feature, from prototype to A/B to rollout.”
    Replace internal codenames with functional descriptions.
    Keep resume to 1 page if under ~8–10 years experience.

  3. Make a focused AI portfolio
    People skim. Build 2–4 strong things, not 10 weak ones.
    Good options right now:
    • An LLM evaluation or RAG project on a real dataset.
    • A small agent system with clear metrics.
    • A recsys or ranking project with offline and online style metrics.
    • Infra work: serving, optimization, monitoring.
    Put it in a GitHub repo plus a 1–2 page PDF “project brief”.
    For each project, show:
    • Problem.
    • Data.
    • Approach and tradeoffs.
    • Metrics.
    • Short code snippets or links.
    Many recruiters do not read code, they read the brief.

  4. Target companies where your Meta experience maps cleanly
    Rough buckets right now:
    • Foundation model labs: OpenAI, Anthropic, Google DeepMind, xAI, Cohere, Mistral. Strong fit if you worked on training infra, evals, safety, LLaMA, distributed systems.
    • Applied LLM product companies: Notion, Figma, Dropbox, Quora, Grammarly, Replit, Salesforce, HubSpot, Monday, etc. They want people who know how to ship LLM features, measure impact, and fix latency and reliability problems.
    • B2B “AI for X” startups: AI for sales, support, code, design, documentation, legal, health. Your Meta brand helps a lot here. They care about speed and ownership.
    • Infra / tooling: weights and biases, LangChain, Modal, Anyscale, Together, MosaicDB, Pinecone, Weaviate, Qdrant, etc. Great if you did infra, observability, or platform.

    Pick 2–3 buckets. Tailor your story to each bucket instead of saying “open to anything in AI”.

  5. Use the layoff to your advantage in outreach
    You have a clean story. Something like:
    “I was part of the recent Meta AI reduction. Looking for roles where I can ship applied LLM or ranking work fast. Here is a quick summary of my background and 3 relevant projects.”
    Most hiring managers have seen multiple Meta or FAANG layoffs. No stigma.
    Post a short, specific LinkedIn post and ask ex coworkers to comment or share. Those posts often get a lot of reach.

  6. Where to find leads right now
    • AI-specific job boards:

    • AIJobs.dev
    • Y Combinator jobs filter “AI”
    • Wellfound (filter by “AI / ML”)
      • Twitter / X: many founders post “we are hiring our first ML engineer” there.
      • Layoff alumni groups: Meta alumni Slack, Blind channels, local tech Slack groups.
      • Your ex managers: ask for 2–3 intros each to founders or hiring managers.

  7. Interview prep for current AI market
    Expect a mix of:
    • Coding: LeetCode medium level is enough for most startup roles.
    • ML system design: “Design an abuse detection system for comments.” “Design a RAG system for internal docs.” Practice clear tradeoffs and metrics.
    • LLM specifics: prompts, context windows, retrieval, embeddings, evals, latency vs quality.
    • Past impact: they care about “what did you personally own” and “what changed because you were there.”
    Keep a single doc with 6–10 “STAR” stories from Meta. Reuse them.

  8. Short AI portfolio ideas you can build in 2–3 weeks
    Pick one that matches what you want next.
    • RAG for support: Ingest product docs, build retrieval, add an eval set of 50–100 real questions, measure accuracy and latency before and after improvements.
    • Ranking project: Use an open dataset like MSLR or a click dataset. Train baseline, then improved model. Show AUC, NDCG, and cost.
    • LLM eval tooling: Build a simple eval harness using GPT or LLaMA as a judge, with guardrails, a dashboard, and seed tasks.
    These show practical thinking, not only model training.

  9. Take care of timing and money
    If you got severance, map your runway.
    Example:
    • Monthly burn.
    • Months of savings plus severance.
    Decide how much time you give to a “perfect fit” search vs “any good offer”. This reduces panic decisions.

  10. Mental side
    Being laid off from a big brand hurts ego and identity. That is normal.
    Set a simple weekly plan to avoid drifting:
    • 10–15 targeted applications.
    • 3–5 warm intros.
    • 1 portfolio improvement.
    • 1 mock interview with a peer.
    Small wins help a lot.

If you share more details, like whether you were more on research, infra, or applied product, people here can give sharper company lists and portfolio ideas.

3

Got hit in the same Meta wave, also on the AI side, so here’s what actually happened for me and a few folks around me, beyond what @sterrenkijker already laid out.

  1. First 2 weeks: don’t “optimize the job search” yet
    I kind of disagree with immediately formalizing your story and blasting apps. I tried that, got burnt out, and my interviews sounded robotic. What helped instead:
  • Week 1: sleep, exercise, talk to 2–3 ex coworkers per day, nothing “formal”
  • Week 2: only low-stakes actions: updating LinkedIn headline, pinging friends, very light resume edits
    Once your nervous system calms down a bit, everything else is easier.
  1. Use your Meta context more aggressively
    You probably shipped or supported something at serious scale. That’s not just resume decor. I started worksing it directly into convos like:
  • “At Meta we learned X breaks at Y scale, so for your product I’d…”
    Recruiters and founders lean in when you move from “I did ML at Meta” to “here’s what scale taught me and how it transfers to your 10x smaller team.”
  1. Portfolio: do less coding, more “thinking in public”
    While @sterrenkijker is right about projects, what actually got me callbacks:
  • One solid project repo
  • Plus 2 or 3 short writeups on:
    • “What I’d change about current LLM eval practices”
    • “How I’d build an MVP RAG system for a 20-person startup”
    • “Here’s how I’d cut LLM infra cost by 30% in 3 steps”
      Post those on LinkedIn or a lightweight blog. Many hiring managers read that more than they read GitHub.
  1. Don’t fetishize only “top AI” companies
    A lot of people tunneled on OpenAI / Anthropic / DeepMind and lost 3–4 months. Some reality:
  • Those roles are insanely competitive
  • They skew research-ish or infra-ish
  • Hiring there often moves very slowly
    Meanwhile, boring sounding companies with real revenue are quietly paying well for applied ML and LLM product work. I ended up with way stronger offers from “AI for X” B2B companies than from the brand names.
  1. Meta layoff signaling: lean into it, but be specific
    Instead of “I was laid off from Meta AI, open to new roles,” I had better results with:
  • “Laid off from Meta AI. Last year I worked on [one sentence plain-English summary]. Now I specifically want:
      1. applied LLM features
      1. ranking / personalization
    • in product-focused teams at startup / growth stage.”
      People forwarded that around because it was easy to understand and pitch.
  1. Warm intros > cold apps, but don’t over romanticize them
    Everyone says “network,” but concrete tactics that worked:
  • Ask every ex manager / tech lead: “Can you intro me to 2 founders or hiring managers where my background is actually relevant?” Make it that specific.
  • Join Meta alumni Slack and literally post:
    • “Former Meta AI [X area]. Looking for [Y kind of role]. Here are 3 bullet points about me.”
      Cold applications still got me about 15–20% of my interviews, so I wouldn’t ignore them, just don’t spend 90% of your time there.
  1. Interview prep: focus on how you think about LLMs right now
    Everyone is memorizing the same RAG diagrams. What stood out in my interviews:
  • Having clear opinions, like:
    • When I wouldn’t use an LLM at all
    • How I’d run cheap but meaningful evals under time pressure
    • How I’d trade latency vs quality for a real feature (e.g. code assistant vs support bot)
  • Having 2 or 3 concrete “failure postmortems” from Meta:
    • “We tried A, it failed because B, here’s what I’d do differently now.”
  1. Emotional / identity side that nobody likes to admit
    There’s a weird shame going from “Meta AI” to “I’m unemployed.” That can distort your search. Two practical tricks:
  • Set a “floor” and “target” job:
    • Floor: roles you would accept if the search stretches too long
    • Target: your ideal scope / comp / tech
      Revisit that every month so you don’t panic-accept too early or chase fantasies too long.
  • Pair up with one other laid off engineer. We literally did a 30 minute check-in every week: sent each other our shortlists, gave each other 1 intro if we could, and reviewed each other’s prep.
  1. Where to aim, given you were on Meta AI
    Rough mapping I’ve seen work well:
  • If you were closer to infra / training / eval:
    • Foundation model labs
    • LLM infra startups
    • “AI platform” teams inside regular companies
  • If you were closer to product / ranking / experimentation:
    • Any SaaS or consumer app trying to ship LLM features fast
    • Growth-stage companies that already have traffic and are trying to optimize engagement or monetization
      The trap is saying “I can do research, infra, product, data, anything.” That just makes you forgettable.

If you’re willing to share 3 things:

  • Were you more research, infra, or applied product?
  • Backend strength level (comfortable owning services, or mostly notebook / modeling)?
  • Geographic flexibility / remote only or not?

Can riff on more targeted company types and 1–2 portfolio projects that will actually match where you want to land, not just “look impressive.”

4

You already got solid tactics from @reveurdenuit and @sterrenkijker, so I will zoom in on a different angle: how to experiment fast without turning the search into a full‑time identity crisis.

1. Treat the next 2–3 months as an exploration sprint, not a linear “job hunt”

Instead of “find perfect job,” structure it as:

  • Hypothesis A: “I want applied LLM / ranking at growth‑stage SaaS”
  • Hypothesis B: “I want infra / platform at an AI tooling company”
  • Hypothesis C: “Maybe I actually like smaller ~Series A teams”

Then:

  • Line up 3–5 conversations per hypothesis (founders, ICs, ex‑colleagues)
  • After ~3 weeks, brutally kill at least one hypothesis

This avoids the trap both of them hint at but do not state: endless “I’m open to infra / product / research” mode.

2. Stop overvaluing “portfolio projects” and undervaluing “reference stories”

Hot take: a lot of AI portfolios look the same now. RAG + vector DB + dashboard. Recruiters are numb.

What is still scarce:

  • Two or three very crisp stories from Meta where you:
    • Changed a metric that mattered
    • Diagnosed a weird failure in prod
    • Pushed back on bad product ideas with data

Put those into a one‑pager that you can attach just like a portfolio. It is faster to read than GitHub and hits harder. Your Meta brand already functions as a kind of product title itself even if you do not explicitly package it as a polished “case study product.”

Pros of this “story sheet” approach:

  • Quick to assemble from work you already did
  • Differentiates you from the RAG side‑project crowd
  • Works well for busy hiring managers who skim

Cons:

  • Harder to anonymize if your Meta work is sensitive
  • Less flashy than a public demo, so weaker for very early founders who want to “see” something

Compared with what @reveurdenuit and @sterrenkijker shared, this shifts focus from “build something new” to “explain what you already did” which is usually a better return on time in the first month.

3. Use “lightweight trials” with companies

Instead of only doing standard loops, suggest small trials when appropriate:

  • 3–5 hour scoped exercise:
    • “I will write a short doc on how to cut your LLM cost 20 percent given XYZ assumptions”
  • Or a 1‑week paid trial if they are open to it

This works especially well with:

  • AI infra startups that need someone who can think clearly about cost and reliability
  • Product teams who are unsure how to scope their first or second AI feature

You are essentially shipping a tiny version of what you would do on the job. That is more convincing than another take‑home ML homework. It also lets you assess if you actually like how they think.

4. Be more opinionated in public than you are in your resume

Both other answers touch on “thinking in public,” but I would push further: hireable AI folks now have opinions, not just checklists of tools.

Pick 2–3 stances and actually write them:

  • “Why most support‑bot RAG systems fail after launch”
  • “When retrieval is overkill and a tuned prompt is enough”
  • “What I learned about guardrails from abuse / integrity work at scale”

Keep each to 600–800 words, post on LinkedIn. No need for a polished blog. A lot of founders read that more than CVs. It also gives you talking points you can reuse in interviews.

5. Categorize your target companies by risk profile, not just by type

Instead of only “foundation labs vs SaaS vs AI for X,” think like this:

  • Low risk

    • Later‑stage public or pre‑IPO with AI teams already in place
    • Pros: stability, resources, mentorship
    • Cons: slower scope growth, more politics

  • Medium risk

    • Series B–D AI for X, real revenue, smallish ML team
    • Pros: big scope, close to product, decent comp
    • Cons: possible reorgs, partial chaos

  • High risk

    • Seed / Series A, “we just raised, need first ML person”
    • Pros: massive ownership, learn everything fast
    • Cons: high failure rate, comp skewed to equity

Map your actual finances and runway to these buckets. Early in severance you can afford to include more medium / high risk. When runway shrinks, deliberately shift the mix of applications instead of just “feeling more desperate.”

6. Do a honest skill audit rather than just claiming “full stack ML”

Quick self check:

  • Can you comfortably design an API, think about SLAs, and own a service?
  • Or are you mainly strong at modeling, analysis, and experimental design?

Both are fine, but the job markets are somewhat different now:

  • Strong backend + decent ML

    • Very attractive for early AI startups
    • You can position as “applied ML + backend engineer” which is rare

  • Strong modeling + experimentation, light backend

    • Great for ML platform, ranking, ads‑like work
    • Less ideal as the “first ML hire” unless they already have solid eng

Write this brutally honest split at the top of your prep doc. It helps you say no faster to misaligned roles.

7. Structure your weeks like a sprint, with explicit “non‑job” time

People underestimate how much context switching kills them. One adjustment that helps:

  • 3 blocks per week for “outbound & interviews”
  • 2 blocks for “portfolio / public thinking”
  • 1 block for “non‑career deep work”
    • Open source contributions
    • Learning a non‑AI skill
    • Or just a personal project

The non‑career block sounds like a luxury but often keeps burnout at bay, which indirectly improves interview performance. I actually disagree a bit with the heavy application quotas some folks suggest. Ten laser‑targeted applications beat thirty generic ones, especially with your background.

8. Use the Meta network, but only after you sharpen your ask

Before pinging ex‑coworkers, finish these two things:

  • A 2–3 sentence blurb of what you want next 2 years, not “someday” career
  • A short list of 10–15 companies or archetypes you care about

Then your message becomes:

“I was affected by the Meta AI layoffs. For the next role I want [X kind of work] at [Y kind of company]. Do you know 1–2 hiring managers or founders where I might be a fit?”

People are much more willing to help when they can immediately think of a couple of names instead of having to figure out your direction for you.

If you reply with whether you were closer to research, infra, or product, and how comfortable you are owning services end‑to‑end, it is possible to suggest a small handful of specific company profiles and one or two “story sheets” you should write first.