▸ CAREER GROWTH · beyond the loop

Career growth — what each level actually demands.

The honest map of the IC and management tracks above Senior. Staff Engineer, Principal Engineer, Engineering Manager, Data Architect, AI Architect — what changes at each step, the L+1 transition signals, and the patterns of work that make the next promotion case obvious. The under-served middle of every engineer's career.

§ 01 — The IC vs management fork

IC vs management — pick consciously.

Most engineers default into management because "that's the next step." It's not. The IC ladder goes to L8/L9 at every FAANG; the work is different. Compensation is identical between IC and EM at the same level. Choose by what you actually want to spend Monday morning doing.

ActivitySenior IC (L5)Staff IC (L6)EM (L6)
Coding~60%~20%~5%
Design / strategy docs~20%~40%~25%
1:1s, hiring, perf~5%~10%~50%
Cross-team / partner~10%~25%~15%
Heads-down focus~5%~5%~5%
The signal you're picking right: on a hard day, what restores you? If it's solving a thorny technical problem alone — IC track. If it's a 1:1 where you helped someone unstick — management track. There's no third answer that survives 5 years.
· · ·
§ 02 — Staff Engineer

Staff Engineer (L6).

L6 · Staff IC · scope of an org

The technical leader of a domain

What's different from L5: you own the patterns, not just the code. You debug across teams. You write the design doc the org refers to next quarter. You're the one peer engineers and engineering managers reach out to BEFORE making an architectural choice.

The four Staff archetypes (Will Larson's framing)

ArchetypeWhat they doWhere they thrive
Tech LeadMost common. Owns a team's technical direction; pairs with the EM.Product teams; healthy mature orgs
ArchitectOwns a long-horizon technical bet across teams.Platform / infra orgs
SolverParachutes into the hardest problem in the org. Moves every quarter.Big orgs with constant fires
Right HandSenior leader's technical thought partner.Director / VP-led orgs

L5 → L6 transition signals

  • You wrote a design doc that another team adopted
  • You ran an incident review where the root cause crossed 3+ teams
  • You mentored an L4 to L5 promotion
  • You said no to a project the VP wanted and were right (not punished)
  • You have a calendar full of being asked for opinions, not being told what to build

The L5 → L6 trap

Being "the best L5 in the org" doesn't promote you to L6. Senior is execution; Staff is influence. The work products are different — design docs, multi-team initiatives, mentorship that shows up in others' promo packets. If you're churning out more code than anyone, you're a great L5. To be L6, write less code and influence more.
· · ·
§ 03 — Principal Engineer

Principal Engineer (L7).

L7 · Principal IC · scope of a BU or company

The technical leader of a domain across the company

What's different from L6: 2-3 year horizon work. You see the curve nobody else sees. You convince VPs to bet on it. You build the team that executes it. Failure is measured in lost competitive position, not lost quarter.

L6 → L7 signals

  • You started a 2-year initiative that's now the org's strategic priority
  • You influenced a roadmap change at the VP / SVP level
  • You're hired into the room before the strategy decision happens
  • You hired or hired-out (managed-out) a Staff engineer
  • External — conference talks, paper, OSS that's adopted

The Principal trap

"I'm the senior-most engineer here" is not Principal. Principal is about strategic insight the org needed and didn't have. If you're answering questions that have been asked before, you're a great Staff. To be Principal, you need to be the one ASKING the strategic question first.
· · ·
§ 04 — Engineering Manager

Engineering Manager.

EM at L5 / L6 / L7

People-leadership track

The EM ladder mirrors IC. L5 EM = first-line manager (5-8 ICs). L6 EM = manages other managers or a senior staff team (10-25). L7 EM = director-equivalent (40+).

First-time EM checklist

  • The IC instinct will betray you. When something's stuck, the IC instinct is to do it yourself. As an EM, that's stealing growth from your reports.
  • 1:1s are the work. Block them, prioritise them, prep for them. Skipping 1:1s for "more important work" is the single biggest EM failure mode.
  • Hire above your level. The biggest leverage is hiring people who will be your peer in 2 years.
  • You will become invisible. Your reports get the credit. Senior leaders measure you on your team's output. Make peace with this.

The IC → EM swing

Most EMs go EM → IC → EM at least once. It's not failure — the work is genuinely different and your taste evolves. Many companies have explicit "swing" support; ask about it before transitioning.

· · ·
§ 05 — Data Architect

Data Architect.

A senior IC role that exists at most enterprise companies (less common at FAANG, which tends to fold it into Staff/Principal DE). The Data Architect:

  • Owns the data platform's long-horizon shape — lakehouse vs warehouse, table format choices, governance model
  • Sets the "blessed path" patterns that product DEs follow
  • Has direct influence on multi-million-dollar vendor choices (Snowflake, Databricks, Confluent, Datadog)
  • Reports up to CTO / VP Eng / Chief Data Officer

Path to Data Architect

  • 5-10 years of senior DE with a track record of platform work, not just product DE
  • A specific "I migrated the company from X to Y" success
  • Vendor relationships — you know the AEs, the product roadmaps, the unreleased features
  • Industry presence — conference talks, blog posts, OSS commits
The Data Architect compensation curve: often higher than the equivalent FAANG L6, because mid-market enterprises pay premium for someone who can de-risk a $10M platform decision. Worth considering even if you've been on the FAANG track.
· · ·
§ 06 — AI Architect

AI Architect.

The 2026 version of the Data Architect role, focused on LLM-native systems. Less established than Data Architect — the title varies (AI Architect, AI Platform Lead, Head of AI Engineering) — but the role is exploding.

What the AI Architect owns

  • The LLM-stack reference architecture — RAG, agents, evaluation, model serving
  • Vendor selection — model providers (OpenAI / Anthropic / open-source), vector DBs, observability
  • The eval pipeline — the rubric every AI product launch must pass
  • Cost attribution and budget — per-product, per-tenant LLM spend
  • The migration plan — what gets prompted, what gets fine-tuned, what gets self-hosted

Path to AI Architect

  • Solid DE / ML engineering foundation (you can't fake the systems work)
  • Hands-on with RAG, agent frameworks, evaluation tools — see the AI Engineering page
  • 1-2 production AI launches you owned end-to-end
  • Strategic voice in the org — you're in the room with the CEO when they ask "what should we do about AI?"
The 2026 AI Architect compensation: top-of-band for senior IC roles. The market is short of people who genuinely understand both production engineering AND LLM-stack tradeoffs. If you have both, you're rare.
· · ·
§ 07 — Manufacturing scope

The L+1 transition — manufacturing scope.

The most common reason engineers stay stuck at Senior: they wait for scope to be assigned. Senior+ engineers manufacture scope. The patterns:

  1. Find the pain — what is the org collectively grumbling about? Tech debt nobody owns, a tool everyone uses but nobody maintains, a metric that's getting worse but isn't on anyone's OKRs.
  2. Write the doc — 1-page proposal. The problem, the proposed fix, the cost. Send it to your skip-level (not your manager — your manager already trusts you).
  3. Recruit a sponsor — get an L7+ to say "yes, this matters, do it." Now you have air cover.
  4. Ship something visible in 4-6 weeks — proves you can execute, not just write docs.
  5. Recruit a partner — bring in another team or function. Now it's cross-team, which signals L+1.
  6. Talk about it — internal demos, all-hands, brown-bags. Senior leaders need to associate the work with your name.
The promotion math. Promotion committees need 3-5 distinct examples of L+1 work. One example is luck; two is a pattern; three+ is undeniable. Manufacture three pieces of L+1 scope over 12-18 months and the promotion case writes itself.
· · ·

Related pages

· · ·