Platform Engineering · Data Infrastructure
Sole Owner
8,000+ students · 3 programs
NUCEE Data Platform:
From Fragmented Silos
to
Repeatable Infrastructure
Northeastern University's Center of Entrepreneurship Education had no
centralized data platform. I built its first: designed the schema,
standardized definitions across three programs, migrated 800+ users to a
new platform, and automated 10+ workflows that had required hours of
manual work each week.
8K+
Students served across VMN, IDEA, and Mosaic
−50%
Reduction in data retrieval time
800+
Mentors and students migrated to new platform
Unified Data Schema · Shared across VMN, IDEA, Mosaic
student_id PKuuid
programenum
cohort_yearinteger
venture_id FKuuid
stageenum
activeboolean
venture_id PKuuid
namestring
stageenum
program FKenum
funding_raiseddecimal
mentor_id FKuuid
engagement_id PKuuid
mentor_id FKuuid
venture_id FKuuid
typeenum
session_datedate
duration_mininteger
Context
Three programs, no shared data model.
The NU Center for Entrepreneurship Education runs three programs: VMN,
IDEA, and Mosaic. Each serves students, ventures, and mentors at
different stages of the entrepreneurship pipeline. When I joined, the
programs operated in silos: separate tracking systems, inconsistent
definitions for the same metrics, and no way to produce a unified view
of NUCEE's impact.
The Princeton Review ranking submission had run the same way for 20
years: manual, fragmented, and dependent on institutional memory
rather than documented process. Every time someone needed a number,
how many active ventures, how many mentor sessions this quarter, what
stage students were at, they had to hunt across multiple tools and
hope the definitions matched.
My job was to build the infrastructure that answers those questions
reliably.
What I Built
Platform design first, automation second.
01
Designed NUCEE's first centralized data architecture
Mapped every data source across the three programs: Salesforce,
Airtable, Chronus, and ad-hoc spreadsheets. Designed a unified
schema to represent students, ventures, mentors, and engagements
consistently. Defined shared field names, data types, and enums
so the same concept meant the same thing in every program.
02
Standardized definitions across three programs, before building
anything
What counts as an "active venture"? What's a mentor "session"?
Each program answered differently. Automating on top of
inconsistent definitions would have locked in the inconsistency.
Aligning on canonical definitions first meant every later
workflow built from the same foundation. This took weeks. The
technical work took days.
03
Built 10+ automated workflows replacing manual processes
Built automations in Airtable and Salesforce to replace manual
data collection, tracking, and reporting: status updates,
reporting triggers, and data sync flows that had required hours
of manual work each week. Cut manual data retrieval time by
roughly 50%.
04
Delivered NUCEE's first unified annual impact report
Produced the organization's first report consolidating data
across all three programs into one narrative: venture counts,
mentor engagement, student reach, and funding outcomes. Used by
200+ stakeholders, including university leadership, program
staff, and external partners.
05
Rebuilt the Princeton Review ranking submission from scratch
Redesigned a process that hadn't changed in 20 years. Documented
the new workflow, identified the required data points, and built
infrastructure to pull them reliably and repeatably. This ended
the manual scramble each submission year.
06
Migrated 800+ mentors and students to a new platform
Directed the data migration and operations redesign for a
NUCEE-affiliated program. Rebuilt mentor and student tracking
for 300+ mentors and 500+ students on a new platform. Scoped the
migration, mapped field relationships, and validated data
fidelity after the migration.
Platform Architecture Decisions
Why these choices, not others.
Salesforce as source of truth, Airtable as working layer
Salesforce held the authoritative contact and venture records, but
program staff worked in spreadsheets, not Salesforce. Airtable
bridged the gap: familiar enough for non-technical staff to use
daily, structured enough to support automations and reporting.
Salesforce stayed the canonical source; Airtable became the
operational surface.
Definitions before tooling, always
The biggest bottleneck wasn't technical. It was semantic. VMN, IDEA,
and Mosaic each had informal, incompatible definitions for shared
concepts. Automating on inconsistent definitions would have locked
in the inconsistency at scale. Aligning on shared definitions first
made every later workflow trustworthy from day one.
Documentation as a platform deliverable
A co-op engagement that leaves behind undocumented systems is a
liability, not an asset. I documented every workflow, schema
decision, and automation: data dictionaries, onboarding guides, and
workflow maps, with enough detail that someone who wasn't there when
it was built can maintain, extend, or debug it. That was a design
constraint from the start, not an afterthought.
Outcomes
A platform that keeps working.
−50%
Reduction in data retrieval time via automated workflows
800+
Users migrated to new platform with validated data fidelity
10+
Manual workflows replaced by automated systems
What I'd Do Differently
What I took from it.
01
Data infrastructure is an organizational problem first
The fragmentation at NUCEE wasn't a technical failure. It was an
organizational one. Three programs grew independently, each with
its own tooling and conventions. Fixing the infrastructure meant
changing how people across the organization thought about shared
data, not just building better pipelines.
02
The most valuable deliverable is the one that keeps working
A report someone recreates manually every year isn't
infrastructure. It's a recurring project. The goal was always to
build systems that produce value after I leave: automation over
one-off analysis, documentation over tribal knowledge, and
repeatability over elegance.
03
Standardization is a leadership challenge, not a technical one
Getting three program teams to agree on shared definitions took
more facilitation than I expected. Building the unified schema
took days. Getting buy-in took weeks. That ratio was right: a
foundation built on contested definitions would have collapsed
the moment someone tried to produce a cross-program report.
Salesforce
Airtable
Tableau
Chronus
Jupyter
Python
Excel
Data Modeling
Process Automation
Stakeholder Reporting