Career Strategy Template¶
A structured framework for career planning during job transitions. Includes the Domain-First Prep process for interview preparation, T-Shape Audit for identifying knowledge gaps, and productivity systems for maintaining momentum.
How to use: 1. Fill in the "Current Direction" section to clarify your path 2. Use the Domain-First Prep process before each interview cycle 3. Run the T-Shape Audit to identify and close knowledge gaps 4. Adopt the MVD (Minimum Viable Discipline) system to maintain consistency
Current Career Direction¶
Decided Path¶
[Your target role type] -- Use first year to [primary goal] -- Reassess at 12 months
Critical Timeline¶
| Deadline | What | Status |
|---|---|---|
| [Date] | [Milestone] | [Status] |
| [Date] | [Milestone] | [Status] |
| [Date] | [Must-have outcome] | [Status] |
Compensation Targets¶
- Minimum base: [Amount]/month ([Amount]/year)
- Target total comp: [Amount]+ (base + bonus + equity)
- Must have: [Critical benefits -- insurance, WFH, etc.]
Current Pipeline¶
| Priority | Company | Why This Role |
|---|---|---|
| 1 | [Company A] | [What moat/capital it builds] |
| 2 | [Company B] | [What moat/capital it builds] |
| 3 | [Company C] | [What moat/capital it builds] |
Three Career Moats (Each Role Type Offers One)¶
| Role Type | Moat Type | What You Become |
|---|---|---|
| Big Tech / High Brand | Technical depth + brand | "[Domain] builder at [Company] scale" |
| Consulting / Varied | Technical breadth | "Advisor who's seen N+ architectures" |
| Domain Specialist | Deep expertise | "[Domain] expert" (transferable within vertical) |
All types are valid. No wrong choice -- each compounds differently over time.
Domain-First Prep Process¶
This is the core interview preparation framework. The key insight: prep should start from the team's business function, not from generic ML study guides.
The Process (Run for Every Interview Pipeline)¶
Step 1: Map the Org's Core Function¶
What does this team/org actually do? Not the JD buzzwords, but the business function.
| Org Type | Business Function | Core Problems |
|---|---|---|
| Sales org | Revenue generation | Forecasting, demand planning, revenue analytics |
| Risk/Compliance | Loss prevention | Anomaly detection, regulatory reporting, explainability |
| Product org | User engagement | A/B testing, recommendation, engagement metrics |
| Ops org | Efficiency | Optimization, logistics, real-time systems |
| Infrastructure | Platform reliability | Scalability, monitoring, deployment |
| Research | Innovation | Novel methods, benchmarks, publications |
Step 2: Derive the Likely Technical Domains¶
What ML/DS techniques serve that function?
| Business Function | Key Technical Domains |
|---|---|
| Revenue/Sales | Time series forecasting, causal inference, MAPE, Prophet, hierarchical models |
| Risk/Compliance | Rare-event classification, explainability, recall-focused metrics, graph ML |
| Product | Experimentation platforms, bandits, ranking models, recommendation |
| Ops | Optimization, routing, real-time inference, constraint satisfaction |
Step 3: Audit Your T-Shape Against Those Domains¶
Can you name the standard tools, metrics, and approaches for each relevant domain? (See T-Shape Audit below.)
Step 4: Close Fundamentals Gaps Before the Interview¶
Even 2-3 hours of focused study converts "can't name it" into "haven't built it but understand the landscape." That's the difference between a concerning gap and an acceptable one.
The meta-lesson: Domain analysis should happen when you start prepping for the company, not the night before. If the JD says "Sales org" -- forecasting prep should start immediately.
T-Shape Audit¶
The T-Shape Concept¶
Senior-level interviewers expect T-shaped knowledge: deep in your specialization, broad enough to speak intelligently about adjacent areas.
| What's Acceptable | What's Concerning |
|---|---|
| "I haven't built forecasting systems, but here's how I'd approach it..." | "I don't have experience with that" + can't name tools/metrics |
| Proactively acknowledges gap + asks good clarifying questions | Gap in both experience AND fundamentals |
| Limited hands-on but understands the landscape and tradeoffs | Can't demonstrate short ramp-up distance |
The key distinction: There's a difference between "I haven't built X" (experience gap -- fine) and "I can't name the standard tools/metrics for X" (fundamentals gap -- concerning). The first shows specialization. The second signals narrow learning habits.
Breadth Audit Template¶
Customize this for your field. Three separate columns track the real state -- having notes does not equal having learned it:
- Coverage: Do you have reference material on this topic?
- Readiness: Can you answer conceptual + technical questions under interview pressure?
- Experience: Hands-on project work (strengthens answers but not required for the T-shape bar)
| Domain | Fundamentals Bar | Coverage | Readiness | Experience | Notes |
|---|---|---|---|---|---|
| [Your core specialty] | [Key tools/metrics] | ||||
| [Adjacent domain 1] | [Key tools/metrics] | ||||
| [Adjacent domain 2] | [Key tools/metrics] | ||||
| [Adjacent domain 3] | [Key tools/metrics] | ||||
| [Adjacent domain 4] | [Key tools/metrics] |
Example for ML/DS roles:
| Domain | Fundamentals Bar | Coverage | Readiness | Experience |
|---|---|---|---|---|
| Classification & rare events | Precision/recall, class imbalance, SMOTE | |||
| NLP / LLMs / GenAI | Transformers, RAG, evaluation, agents | |||
| Time series forecasting | ARIMA, Prophet, MAPE, hierarchical models | |||
| Recommendation systems | Collaborative filtering, embeddings, ranking | |||
| Causal inference | A/B testing, DiD, uplift modeling | |||
| MLOps / production | CI/CD, drift, monitoring, deployment | |||
| Experimentation | Statistical significance, power analysis, sequential testing |
Readiness Pipeline¶
The honest path from gap to battle-tested:
| Level | What It Means | Can You... |
|---|---|---|
| No material | No reference material exists | -- |
| Content added | Material captured, not yet studied | Look it up if asked |
| Read through | Studied the material once | Recognize terms, explain at high level |
| Q&A drill | Practiced with mock questions | Answer conceptual + technical questions under pressure |
| Interview-ready | Can discuss fluently without notes | Name tools, articulate tradeoffs, connect to experience |
| Project-proven | Built something end-to-end | Share war stories, debug edge cases, explain implementation decisions |
Each step is distinct. Don't skip steps or mark something ready just because notes exist. Project-proven is the gold standard -- it separates "I've read about X" from "I've used X in production and here's what surprised me."
Using the Audit¶
Before a new interview pipeline: Run Domain-First Prep (above) to identify which domains matter, then check readiness for those domains.
Between interviews: Pick one gap per week to move through the readiness pipeline.
Goal: All domains at minimum -- material exists + reviewed + can articulate standard tools and tradeoffs under pressure.
Interview Lessons Framework¶
LLM Self-Assessment Calibration¶
If you use LLMs to analyze interview transcripts or mock interviews, be aware of systematic biases:
| LLM Bias | What Happens | Correction |
|---|---|---|
| Social signal over-weighting | Positive comments and engagement read as pass signals | These are politeness norms, not evaluation signals |
| Domain-agnostic scoring | Scores breadth equally regardless of role focus | Weight gaps in the role's core domain 2-3x higher |
| Compensatory assumption | Assumes strength in area A compensates for gap in area B | Interviewers evaluate each competency against a bar, not on average |
| Candidate perspective blindness | Dismisses candidate's gut feeling as "underestimation" | You were in the room -- your read on interviewer reactions has signal |
Rule of thumb: When LLM estimate diverges significantly from your self-assessment, your self-assessment is probably more calibrated -- you were in the room, the LLM only read a transcript.
Productivity Systems¶
Minimum Viable Discipline (MVD)¶
The core insight: consistency beats intensity. A system that survives bad days is better than one that only works on good days.
Bad Day Baseline (10-20 min) -- "Keep the chain alive"¶
- Add 3 lines to documentation
- Send one outreach message
- Practice one interview story out loud
Normal Day (60-120 min)¶
- 10 min: choose 1 outcome
- 45-70 min: build/write
- 10 min: document/commit
High-Energy Day (deep work)¶
- 2 x 45-50 min focused blocks
- No "catch up" -- high-energy is bonus, not debt repayment
Traffic Light Tracking (Not Streaks)¶
Streaks create shame spirals when broken. Use traffic light tracking instead:
- Green: Full session (60-120 min)
- Amber: Bad-day baseline (10-20 min) -- counts as a win
- Red: Nothing
Goal: Reduce red runs, not maximize greens.
Relapse Plan (After Missing 2-3 Days)¶
- Say: "I'm restarting, not catching up."
- Do the smallest task (10 minutes)
- Log a win ("Returned")
- Pre-load tomorrow: open the repo + leave a note for next step
Success metric: Restart speed, not streak length.
The Core Loop to Break¶
Rigid/unrealistic targets --> miss --> shame --> avoidance --> inconsistency
This pattern resurfaces whenever: - "Learn everything" becomes the new rigid target - Rejections trigger shame spirals - New job pressure exceeds capacity
Recognition is the first step to breaking the loop.
High-Leverage Habits¶
- Sleep protection -- Fixed caffeine cutoff, phone out of reach, wind-down routine
- Anti-shame tracking -- Traffic light system (amber counts as a win)
- Environment design -- Keep a "NEXT" file with the next 3 micro-steps. Reduce friction to start.
- Health-first days -- On bad health days, do bad-day baseline only. No planning, no guilt.
- Weekly closure ritual (20 min) -- Choose next week's 3 outcomes, delete the rest
Useful Scripts¶
30-Second Pitch¶
"I'm a [role] who's strongest at [core strength -- what makes you different]. Over the past [timeframe] I've [1-2 key accomplishments]. I'm looking for [what you want next] at a company where [what matters to you]."
"Why Are You Leaving?" Answer¶
"After [X years] at [company], I've grown from [starting point] to [current level/scope]. I'm looking for [what's next -- depth, breadth, domain, scale] at a company where [specific draw]."
Boundary Script (Protecting Productive Time)¶
"Given current constraints, I'm focusing on one deliverable at a time. If priorities change, I can swap -- but I can't stack."
Decision Frameworks¶
The "Impact vs. Cutting-Edge" Tension¶
Many people feel pulled between (a) state-of-the-art technology and (b) impact-driven work. This is often a false dichotomy.
What actually sustains motivation (look at your own evidence):
| Experience | Sustained? | Why? |
|---|---|---|
| [Past role 1] | Yes/No | [What about the daily work energized or drained you] |
| [Past role 2] | Yes/No | [What about the daily work energized or drained you] |
| [Current situation] | Yes/No | [What about the daily work energizes or drains you] |
Bottom line: Motivation that depends on a narrative ("I'm saving the world" / "I'm doing cutting-edge AI") fades. Motivation from the actual daily work -- shipping, iterating, seeing results -- sustains. Pick the role where the daily work energizes you, and let the narrative follow.
Regret Asymmetry Analysis¶
When choosing between options, ask: which choice is easier to recover from if it turns out wrong?
| Scenario | Regret Level | Recovery Path |
|---|---|---|
| Take Option A, miss Option B's niche? | [High/Low] | [Can you get back to B later?] |
| Take Option B, miss Option A's advantages? | [High/Low] | [Can you get back to A later?] |
One option's value is usually more guaranteed (e.g., brand + comp are structural). The other's value is more conditional (e.g., depends on team, manager, scope). Factor this asymmetry into your decision.
Key Quotes¶
"Outperforming in a weak org buys you internal power, not long-term freedom."
"It's not 10 years of any work. It's 10 years of compounding the right capital."
"Career capital doesn't compound from being the best at one thing. It compounds from sitting at a rare intersection."
"Most people overestimate what they can do in one year and underestimate what they can do in ten years."
Review this template at the start of each job search cycle. Update with lessons learned after each interview round.