╔══════════════════════════════════════════════════════════════╗
║ Muhammad Taqi Haider — Backend Engineer ║
║ Performance Obsessive · BS CS Graduate @ NED ║
╚══════════════════════════════════════════════════════════════╝
I get a weird kind of satisfaction when a slow API becomes fast.
name: Muhammad Taqi Haider
role: Associate Software Engineer @ QBS Co. # promoted within 3 months
education: BS Computer Science — NED University · Graduated June 2026 (GPA: 3.32)
status: Full-time engineer · Available for remote & relocation opportunities
focus: Backend Performance · Microservices Architecture · ML-Backend Integration
exploring: System Design Fundamentals · Clean Architecture · Where AI meets real engineering
open_to: Remote & relocation opportunities globally
contact: taqihaider591@gmail.comGoal: Engineer backend systems that perform under real load — not just in theory. I care about latency, correctness, and systems that don't page you at 3am.
Numbers from real systems. Not estimates.
| What I Shipped | Stack | Result |
|---|---|---|
| Redis caching on high-traffic monitoring endpoints | NestJS · Redis | 1.6s → <600ms — 63% drop |
| Refactored service-layer DB calls across endpoints | PostgreSQL · NestJS | 6–8 → 2–3 hits/request — 60% reduction |
| Rebuilt sync PDF pipeline to async | Azure Queue · SignalR | ~1.6s → 400–500ms — 70% faster |
| PostGIS spatial indexing on multi-tenant SaaS | PostgreSQL · PostGIS | Geospatial queries +25% faster |
| SBERT + Logistic Regression categorisation model | Flask · Hugging Face | 88% accuracy in production |
track-wise2.vercel.app · Final Year Project · Group of 3
AI-powered personal finance platform. Designed, built, and deployed end-to-end.
Backend → NestJS · PostgreSQL · 56 endpoints
ML → Flask microservice on Hugging Face Spaces (3 endpoints)
Frontend → Angular PWA
Three models in production:
| Model | Approach | Result |
|---|---|---|
| Transaction Categorisation | SBERT + Logistic Regression | 88% accuracy — surfaced at entry time |
| Expense Forecasting | Facebook Prophet | Next-month prediction from 3+ months history |
| Anomaly Detection | Custom classifier | AUC-ROC: 0.87 — real-time via API response |
Engineering decisions worth noting:
- ML inference flows through NestJS — Angular never touches the ML service directly
- Response caching + rate limiting to handle Hugging Face cold start latency
- Async event-driven email workflows for account verification and password reset
- Real-time anomaly alerts delivered in API response, rendered by Angular at runtime
Daily drivers:
Used in production:
Patterns I reach for:
Algorithmic challenges as a way to sharpen fundamentals — not just to pass interviews.
Categories covered →
Data Structures
Array Linked List Stack Queue Binary Tree BST Hash Table Heap Matrix
Algorithms
Binary Search BFS DFS Dynamic Programming Greedy Divide & Conquer
Backtracking Two Pointers Sliding Window Prefix Sum Bit Manipulation Graph
Database
SQL 50 (complete) Joins Aggregations Subqueries Schema Optimisation
If you're building something where performance and architecture matter — let's talk.