Project archive: systems engineering case studies

A collection of product engineering case studies: cloud consoles, realtime systems, and Web3 platforms by Sangeet Banerjee.

Back to Sangeet Banerjee homepage

Datareum: project facts

Project
Datareum
Period
2022
Role
Team Lead · Fullstack Engineer
Summary
Blockchain hospital data exchange platform
Description
A healthcare data platform for storing and sharing anonymized hospital datasets with researchers. I led the team building the product around secure uploads, a data-picking pipeline, passwordless auth, Ethereum smart contracts for access control, and a donation flow for institutions contributing data.
Domains
Healthcare, Web3, Data Privacy, Research Systems
Technologies
Next.js, TypeScript, Ethereum, Smart Contracts, Machine Learning
Ownership
Team leadership across frontend and product delivery
Ownership
Data-picking and upload flows for hospital datasets
Ownership
Passwordless OTP authentication experience
Ownership
Ethereum smart contract integration for access and donations
Ownership
Dataset browsing UI with 51+ selectable attributes
Ownership
Developer documentation and onboarding content
Learning
Healthcare and blockchain together force you to be painfully explicit in the UI. Users need to see exactly what data leaves the hospital, who can access it, and what a wallet signature actually means.
Professional signal
I can lead teams through domain-heavy products where security, compliance, and UX all have to show up in the interface.

Datareum

blockchain hospital data exchange platform

Datareum
Overview

a healthcare data platform for storing and sharing anonymized hospital datasets with researchers. i led the team building the product around secure uploads, a data-picking pipeline, passwordless auth, Ethereum smart contracts for access control, and a donation flow for institutions contributing data.

Why this existed

hospitals needed a way to share research-ready data without losing control of who could access it or how keys were managed.

How I approached it

i helped design a pipeline that used machine learning to extract relevant rows from uploaded CSVs, applied SHA-256 for sensitive handling, and used Ethereum smart contracts for transparent key management. on the product side we built passwordless OTP login, dataset browsing with dozens of attributes, Ethereum-based donations, and developer documentation so others could integrate.

What changed

we delivered a complete platform on testnet and production demos that showed how hospital data could move between providers and researchers with clearer rules. it was my main work during my time at Datareum as a fullstack engineer and team lead.

Year2022
RoleTeam Lead · Fullstack Engineer
RoleTeam Lead
Attributes51+ Data Fields
What I actually owned
  • ·team leadership across frontend and product delivery
  • ·data-picking and upload flows for hospital datasets
  • ·passwordless OTP authentication experience
  • ·ethereum smart contract integration for access and donations
  • ·dataset browsing UI with 51+ selectable attributes
  • ·developer documentation and onboarding content
What became difficult

we were juggling privacy, blockchain trust, and messy real-world CSV files at the same time. the extraction algorithm had to pull the right fields from inconsistent datasets, encryption had to feel solid, and researchers still needed a UI that did not require a PhD in Web3 to request access.

What I learned

healthcare and blockchain together force you to be painfully explicit in the UI. users need to see exactly what data leaves the hospital, who can access it, and what a wallet signature actually means.

What surprised me

the hardest bugs were not in the contracts. they were in normalizing CSV files that every hospital formatted differently.

Before → After
Before
  • ·opaque hospital data silos
  • ·manual CSV wrangling
  • ·unclear access control
After
  • ·anonymized shared datasets
  • ·ML-assisted extraction
  • ·contract-backed key management
Small opinion

trust in data products is mostly about clarity, not buzzwords.

What I would improve now

i would move heavy extraction to a background job queue with clearer progress states, and simplify the researcher request flow so legal review steps are visible instead of hidden in modals.

Random things I remember
  • ·we whiteboarded the trade form-style flows for data requests more times than I expected for a research product.
  • ·seeing the first successful OTP login without passwords felt like magic compared to our older auth attempts.
What this project says about me

i can lead teams through domain-heavy products where security, compliance, and UX all have to show up in the interface.

Built with
Next.js
TypeScript
Ethereum
Smart Contracts
Machine Learning
Domains
Healthcare
Web3
Data Privacy
Research Systems