if you do what everyone else does, you will not get hired.

1:1 mentorship to land your first data job

I'm Mark, a senior analytics engineer. I help students and early-career people build the profile a recruiter can't scroll past: projects nobody else has, a GitHub that proves your resume, a story only you can tell.

add me on LinkedIn with a note

A row of identical candidate profiles, and one highlighted profile that is different

why me

In the final round for my data engineering internship at Lazard, I was up against Ivy League students. The offer email went to five .edu addresses — Penn, Brown, Cornell, Columbia, Columbia — and one personal Gmail. Mine. TCU isn't a target school.

The offer email's recipient list: five .edu addresses from Penn, Brown, Cornell and Columbia, and one Gmail address, highlighted — mine

So how did I get the same offer? Not the school.

you won't out-school them. you can out-prepare them.

read the full story in the playbook

people I've taught

“He genuinely believes in his students and encourages us to think deeper instead of simply giving us the answers. He challenges us to be curious, ask better questions, and grow into more confident data professionals.”

Mia Tran — SDS & Quant Econ @ Smith College

“Mark is the kind of mentor that will push you forward and is very critical of your work — he will not let you look back on failures without lessons attached. I believe working with Mark is the fastest way to level up your skills in a corporate setting.”

Gwen Nguyen — Cloud & Data Engineer Intern @ Arch Insurance

“He took me from 0 to 1 — teaching me dbt, Airflow, and GitHub, and walking me through my first PR. As a business analyst with no prior data engineering experience, it was one of the most impactful months since I started working.”

“I've learned more technical skills from him than from everyone else combined.”

what my mentees built

One real pipeline, built together: 117,000 airline reviews scraped from Skytrax, staged on S3, loaded into Snowflake and modelled in dbt, run by Airflow and shipped with Terraform. Then each mentee took one airline and shipped a dashboard on top of it.

The skytrax_reviews pipeline: AirlineQuality reviews scraped and cleaned in Python, staged on S3, loaded into Snowflake with COPY INTO and a load audit, modelled in dbt, then four mentee dashboards on top, highlighted. Airflow runs it; Terraform, GitHub Actions and OIDC provision and deploy it
skytrax_reviews: from scrape to dashboard
Average rating per carrier: industry 2.59 with 40% recommending, Delta 2.49 and 29%, Spirit 1.59 and 12.1%, Frontier 1.43 and 5.7%, highlighted as the lowest
what the dashboards found: Spirit and Frontier sit far below the industry
A hand-drawn dbt lineage: stg, int_cleaned, three dims, fct_review highlighted and tested one row per review_id, then agg_airlines and the Spirit dashboard

skytrax_reviews

Mark + mentees

The shared pipeline: scraper, S3, Snowflake, a dbt star schema, Terraform and slim CI/CD.

A hand-drawn table of how each factor correlates with the average rating for Delta, economy against non-economy; ground service, 0.88 against 0.83, is the widest gap

airline_customer_exp_analysis

Alyssa Le

Delta: why Economy and Premium passengers are happy for different reasons.

Hand-drawn bars of cabin staff and wifi scores by seat type; first class cabin staff is highest at 2.6

frontier-reviews-dashboard

Gwen Nguyen

Frontier against its ultra-low-cost peers.

Hand-drawn bars of Spirit average rating by seat type: first 2.103, economy 1.577, premium economy 1.564, business 1.286

spirit_airlines_dashboard

Mia Tran

Spirit: a deep dive into chronic dissatisfaction.

A hand-drawn star schema: fct_airline_reviews in the middle, joined to dim_date, dim_airline, dim_cabin, dim_country and dim_traveller_type

Skytrax-local-warehouse

Tony Dam

The same reviews on one laptop: DuckDB, dbt and Streamlit, no cloud.

A hand-drawn sketch of the Flightly explorer: a sidebar of sections, an airline picker, three metric cards and a rating-over-time chart

skytrax_reviews_dashboard

Tien Nguyen

A Next.js explorer over the same data.

sound familiar?

MySQL or Postgres? Power BI or Tableau? pip or uv? Everyone starting out freezes at the same tool choices — and none of them is the real question. The principles transfer; we pick one stack and move.

MySQL, PostgreSQL, Oracle, SQL Server — which one do I learn? Just learn ANSI SQL
which SQL? a query is a query
The BI tool changes every few years; the principles — accurate, clear, empowering, succinct — don't
Power BI or Tableau? wrong question
pip for quick scripts, Poetry for long-lived projects, uv for speed
pip, Poetry or uv? each has one job

I've been exactly here. I struggled hard, with no direction.

I built this so you don't make the same mistake.

who it's for

Students and early-career people aiming at one of three tracks. You don't need experience: you need reps.

data analyst

SQL, dashboards people actually use, and analysis that answers a business question nobody asked you to answer.

analytics engineer

dbt, warehouses, tested models, and pipelines that turn raw data into decisions.

data engineer

Ingestion, orchestration, streaming, and infrastructure that keeps running after you log off.

DE hunts and hauls the data in, AE cooks it, DA and DS run the dining room
where each track sits in the kitchen

what we work on

Everything comes back to two questions. The copy-paste test: could 1,000 other candidates paste this line unchanged? The only-me test: what is true about you that isn't true about them?

The whole playbook is public on GitHub. The mentorship is where we apply it to you.

rules you'll hear a lot

what I'll teach you

From the fundamentals to Spark: the SQL that solves 80% of problems, dbt and the modern warehouse, ingestion pipelines with zero infrastructure, and Spark for when the data outgrows one machine. It's all written down in the handbook — the mentorship is applying it to your project.

SQL roadmap staircase from beginner to proficient; intermediate solves 80 percent of problems
sql 101: a roadmap, not a course list
Raw data lands in the warehouse from Salesforce, HubSpot, Postgres; dbt is the T in ELT
dbt 100: the T in ELT
Neon Postgres to Snowflake through the WAL, CDC and Fivetran — read the log, not the tables
pipelines 101: zero infrastructure
seven more: projects, resume, laptop setup, dbt 101 to 302, Spark
A data project folder structure where each folder has one job
projects that don't look like everyone else's
Resume files named final, final v2, final real, versus a clean git history
a resume treated like a product
Four laptops set up by hand, all different — versus one repo and four scripts, about 20 minutes
a laptop that sets itself up
Ad-hoc SQL scripts with no clear order, version control or testing; dbt run builds source, staging model, mart model — parents first
dbt 101: parents first
A table model rebuilds all rows every single run; an incremental model adds only new rows where updated_at is past the max
dbt 202: only the new rows
amount / 100 copy-pasted across three models, versus one cents_to_dollars macro feeding all three — change it once
dbt 302: change it once
Spark execution hierarchy: application, job, stage, task; four 128MB partitions feed executor cores — 1 partition = 1 task
spark: 1 partition = 1 task

things I built that you can use today

I don't only talk about projects nobody else has. These are mine, and they're free.

qdbd

Turns QuickDBD, DBML or a dbt project into an ER diagram that stays readable at 200+ tables. 40 releases, 21 live demos.

CTE Diff

A VS Code extension: preview any dbt CTE and compare it on both sides of a diff. Try it in the browser playground.

dbt-Fundamental

One dbt project that runs on DuckDB or Snowflake. 40 stars, 12 forks.

dbt CI/CD course

Nine hands-on lessons, each with a checkpoint.

how it works

  1. add me with a noteConnect on LinkedIn and tell me who you are, which track you want, and where you're stuck.
  2. we talkA 30-minute intro call. If we're a fit, you join the mentee Slack.
  3. you do the repsYou work through the guides, build, and ship. Spotter, our Slack bot, posts everyone's GitHub commits daily — the group sees your reps.
  4. I review the hard partsSend work only after the checklist passes. Sessions go to the hard parts, not typos.
A commit timeline growing day by day into a chart trending up

about Mark

Senior Analytics Engineer at Infinite Lambda — the youngest L6 there — and ex-Insurify and Lazard. I did three internships in two countries before my senior year, and I've been mentoring since 2020: a small academy that helped 20 Vietnamese students apply to US universities, the AE Academy for new hires at Insurify, and Ace Mentorships. I moderate Xóm Data, Vietnam's largest data learning community (110,000+ members), and write The Analytics Engineer's Handbook on Substack. I mentor because teaching is how I learn.

A sketched portrait in a frame, labelled mark

faq

do I need experience?

No. Most mentees are students or career-switchers. You need time to do the reps, not a data job.

what do sessions look like?

1:1 reviews of your profile, resume, portfolio, and projects — after you've run the checklists yourself.

how often do we meet?

Weekly, or more often when you need it.

how many mentees do you take?

Up to 10 at a time.

English or Vietnamese?

Both. The guides are in English; we talk in whichever you think in.

what happens after I get an offer?

You stay in the Slack. You keep access, and you can help the next mentees.

what does it cost, and how long does it take?

We go through cost, how long it takes and how many hours a week in our first 30-minute meeting.

how do I join?

Add me on LinkedIn with a note. That note is your first copy-paste test.

the note is the application

A LinkedIn connection note: who you are, target role, GitHub link, where you are stuck

add me on LinkedIn with a note