Diana Bârsan
Principal Software Engineer
Summary
Principal engineer with 17+ years of experience, nearly 9 at Medic. There I architected and owned four iterations (2018–2026) of the data replication behind the open-source Community Health Toolkit (CHT), which keeps health workers' phones working without internet, as single deployments grew from a few hundred to 20,000 users. The platform serves 200,000+ community health workers (CHWs) in seven countries. #1 contributor since 2018 to its public codebase:
github.com/medic/cht-core. Deep expertise in performance, scalability, and distributed data. Fully remote since 2009.
Experience
Principal Software Engineer — Medic (formerly Medic Mobile) · Remote
Jan 2022 – Sep 2026
- Rearchitected data replication with a custom endpoint in place of CouchDB's built-in protocol (2023), cutting replication time by 90%; added monitoring with failure tracking
- Migrated replication queries to a faster search engine (2025), shaving a further 60% off replication time and 50% off system load in production; together, single deployments scaled from a few hundred to 20,000 active users
- Built a pipeline archiving stale records to cold storage, reducing median replication time from 2.8 minutes to 6 seconds and 1-hour timeouts to 3 minutes in one production deployment; admins adopted it the next day
- Set the technical direction for a team of about six and defined the long-term platform strategy; ran several open-source squads of outside contributors and mentored 10 of them, some still contributing today
- Unblocked a year-stalled analytics-pipeline replacement in 3 weeks by swapping an unmaintained, data-loss-prone stage for a custom connector; now runs analytics for every deployment
- Led a year-long, LLM-informed UX overhaul of health workers' task workflow (offline prioritization, search, filters, GPS capture): partners reported 30% higher task completion six months after release
- Engineered automated performance benchmarks (EC2, database endpoints, frontend) that gate every release: caught a 20x indexing slowdown in an upstream database release, which I reported and upgraded after the fix; also measured the replication gains above
- Cut server load in deployments with aggressive data purging: a caching layer for retention checks removed redundant requests that took up to 40% of server time
- Architected the move from a monolith to Docker services with one-click, zero-downtime upgrades; Docker by default for teams without ops staff, Helm charts for Kubernetes where scale demanded it
- Fixed a contractually required audit trail for healthcare deployments: replaced truncated proxy request logs with an audit database recording every document change, its user, and request ID
Software Engineer — Medic Mobile · Remote
Jan 2018 – Dec 2021
- Rewrote the CHT's replication engine (2018), the foundation of the platform's offline-first replication; previously, users with too many documents could crash the server
- Introduced per-record access control, closing a gap where users had near-unrestricted API access: limited users to the app's own endpoints and enforced permissions on every write
- Implemented server-side data-retention controls (2019) and configurable limits on how much data each user downloads, making the app usable for users overseeing many health workers, who had abandoned it as too slow
- Drove the full rewrite of the web client from AngularJS to Angular 10; fixed a 2–3x page-load slowdown, caused by the framework re-checking the page on every background data operation, before release
- Integrated SMS/messaging gateways and added token login: users get an SMS link that signs them in, replacing plain-text credentials circulating in spreadsheets
- First responder for production incidents and live deployments; stopped an SMS auto-reply loop that caused an outage in a telecom's network in Nepal, by adding loop detection
Full Stack Web Developer — Novisites VOF · Netherlands
Aug 2012 – Aug 2017
Developed and maintained custom web applications and Prestashop webshops that ran in production for years: integrated Amazon, payment processors (PayPal, Klarna) and shipping platforms; engineered a shared cart across a platform's separate webshops; took over a popular legacy shop (~1 order every 2 minutes) generating personalised printable PDF gifts
Full Stack Web Developer — Web One Division / Eclipse Concept · Romania
Mar 2010 – Jul 2012
Developed Magento webshops, CMS, and business applications with payment integrations; created an interactive news treemap sizing articles by importance and hits
Earlier: Freelance Web Developer (2009–2010) · Intern, Produweb, Belgium (2008–2009)
Open Source & Community
- #1 contributor to cht-core since joining in 2018; #4 all-time on a codebase dating to 2013 (143 contributors) — full history public at github.com/dianabarsan
- Rewrote the technical and architecture documentation for the CHT (docs.communityhealthtoolkit.org)
- Technical support to CHT implementers on the community forum — 1,100+ posts since 2019
- Upstream open-source contributions, including PouchDB, CouchDB, WebdriverIO
- Author of CHT release announcements and technical posts at medic.org
- Side projects: custom game modules used by 50,000+ players; an AI-assisted scraper that scores real-estate listings against personal criteria
Technical Skills
Languages: JavaScript, TypeScript, Python, PHP, Bash
Backend & Data: Node.js, CouchDB, PostgreSQL, MySQL, Redis, Express, REST APIs, dbt (data build tool)
Frontend: Angular, React
Infrastructure & DevOps: Docker, Kubernetes, AWS, Nginx, HAProxy, CI/CD (GitHub Actions), Grafana, Prometheus, Linux
Architecture & Workflows: Offline-First Systems, Distributed Data, Performance Benchmarking, AI-Assisted Engineering (Claude Code daily workflow, custom Claude Code skills, AI-assisted code review, LLM analysis of usage data, local-model scraper)
Languages
Romanian (native) · English (fluent) · French (intermediate) · Spanish (beginner)
Education
BSc, Computer Science — West University of Timișoara, Romania
2004 – 2008