Eve Healthcare
Software Engineer Intern
- Search Optimization: Improved platform search efficiency, lifting user search success rates by 15%.
- Real-Time Communication: Added real-time in-app chat over WebSocket, lifting user engagement and retention by 30%.
- API Performance: Optimized 22% of patient dashboard APIs, cutting database queries and reducing response times by 1 second.
- Doctor Dashboard: Built the doctor dashboard and analytics surface, improving data access by 20% and data-driven decisions by 15%.
The story
idea → framing → build → deploy → what's next01 IDEA
Search Optimization: Improved platform search efficiency, lifting user
Eve Healthcare was a three-month internship during my final year of MCA. The product was a doctor-patient platform with search, in-app chat, and a doctor-facing analytics dashboard. I worked on the backend surface that powered all three.
02 FRAMING
The search work was the first time I touched Elasticsearch end-to-end.
The search work was the first time I touched Elasticsearch end-to-end. The default analyzer was tokenizing English like a search engine built for the late 90s, returning results that matched keywords but missed synonyms and partial matches. I rewrote the analyzer, added synonyms from a small medical dictionary, and tuned the fuzziness parameter. Search success rate went up 15% in the next month's metrics.
03 BUILD
The in-app chat was a WebSocket layer over Django Channels.
The in-app chat was a WebSocket layer over Django Channels. The hard part wasn't the socket itself; it was backpressure when a patient's connection dropped mid-message and the unread-counter had to be authoritative. I learned the offline-first lesson here — never assume the socket is up.
04 DEPLOY
The dashboard work was the cleanest deliverable.
The dashboard work was the cleanest deliverable. Twenty-two percent of the patient dashboard APIs had avoidable database round-trips (N+1 patterns, missing select_related). One refactor pass, latency dropped by a full second on the dashboard hot path. That fix taught me the lesson NexBell then made me formalize: profile and fix the slowest query path before writing a single feature line.
05 WHAT'S NEXT
Outcome — Search Optimization: Improved platform search efficiency, li
The whole internship was when I internalized that production work is mostly about choosing what not to ship. We had five candidate features in the queue and we shipped two of them well rather than four of them badly.
Keywords
3 techs · links go to /stackRelated writing
2 posts reference Eve Healthcare- Your Easy Guide to Design Highly Available Systems With Redis. Part 1
medium · 8 min read · redis · high-availability · cache · distributed
- Step-by-Step Guide to Deploy Redis Cluster in AWS to Achieve High Availability. Part 2
medium · 10 min read · redis · aws · cluster · high-availability