Production systems, open-source tools, and published research. Scope and ownership
stated plainly — no invented metrics.
Healthcare HL7 Integration Platform (under NDA)
Production full-stack work on a commercial HL7 healthcare integration engine at TransData (Jan–Mar 2026), where I was the top contributor (~248 commits). Product name withheld under NDA. Built a React 19 + TypeScript + Vite frontend and a Node.js + TypeScript Express-style BFF (Sequelize, Jest) proxying the platform's server API. Delivered PHI-access auditing features, real-time message-trends dashboards, and accessibility work (ARIA, focus management, keyboard navigation). Operations tooling for healthcare integration teams — scope and ownership stated plainly; no clinical-software or compliance-ownership claims.
Top contributor (~248 commits)
React 19 + TS frontend
Node.js + TS BFF
React 19 TypeScript Node.js Express Sequelize Jest Vite HL7
HireTrack — AI Résumé Builder & Job Pipeline Tracker
Open-source, bring-your-own-key web app that tailors ATS-ready résumés to each job and tracks an interview pipeline — the tool I use to run my own job search. A deterministic-first LangGraph pipeline extracts job-description data and scores match/positioning; résumés export to ATS-safe DOCX, PDF, or Markdown entirely client-side. Truth-only tailoring mirrors JD phrasing with no invented skills or metrics, and every run appends honesty/verification notes. Supports 5 AI providers (Anthropic, OpenAI, Gemini, and any OpenAI-compatible endpoint), Supabase Postgres + Auth with row-level security, and installs as an offline-first PWA.
Primary DRF backend developer for a production AI-powered LinkedIn content platform. Built 7 backend services and 60+ REST endpoints, including versioned prompt management across AI content pipelines (n8n), multi-tier profile extraction (LinkedIn API → scraper → OpenAI fallback), real-time notifications via Django Channels WebSockets, Stripe-based subscription billing, OAuth2 + JWT refresh auth, and semantic search using Gemini embeddings + pgvector with Redis caching.
Architected a real-time intelligence backend as sole engineer, ingesting social data from Telegram, Discord, and X and processing it with Google Gemini embeddings and pgvector semantic search. Built async multi-chain blockchain analysis services with token-bucket rate limiting and Redis caching, plus ML-based DEX scam-token detection. Orchestrated ~10 services via Docker Compose with migrations, background workers, and structured logging.
~10 microservices
6 chains analyzed
~92% scam detection (test set)
Python FastAPI Docker Google Gemini pgvector Redis Celery ARQ Etherscan
Enterprise HRMS Migration & Automation
Migrated and extended a Django-based HRMS (Horilla) for a ~200-employee workforce with full backend ownership. Built attendance, contract, and reporting modules, implemented ZKTeco biometric sync via Web API, designed a multi-device biometric schema, and built monthly/day worksheet reporting endpoints. Deployed a Dockerized system on DigitalOcean with Nginx, Celery background jobs, structured logging, and zero-downtime rolling deploys.
Customized ERPNext / Frappe v16 with custom doctypes, server scripts, client scripts, and page-level UI — including an Asana-style task-tree view for internal task management. Used the bench CLI and REST APIs for programmatic data management and bulk fixes, and researched ERPNext authentication/SSO options for integrating a custom Parts-Intelligence dashboard with ERP-backed access. Worked through role/permission, private-file, and reverse-proxy/WebSocket production concerns.
Built and debugged a web interface for healthcare HL7 integration operations, replicating key Mirth Connect Java-client workflows as a browser-accessible tool. Implemented operational features including message export, channel start/stop, deploy/undeploy, and column drag-to-reorder, plus authentication workflows. Produced Jira breakdowns for channel-operation features.
Mirth Connect web tooling
Channel operations
HL7 message export
Python HL7 Mirth Connect Healthcare Integration Web UI Auth
Local LLM Inference Infrastructure
Deployed internal LLM inference for team use: vLLM on an RTX 3080 Ti (12 GB VRAM), with a LiteLLM proxy exposing a unified OpenAI-compatible API. Benchmarked local models under hardware constraints and selected Qwen3-8B-AWQ for coding and automation workflows, tuning context-window and prefix-caching settings. Wired OpenCode for agentic coding.
vLLM on 12GB VRAM
LiteLLM OpenAI-compatible proxy
Local model benchmarking
vLLM LiteLLM Qwen3-8B-AWQ Python OpenAI-compatible API GPU Inference
Enterprise AI/ML Pipeline Integration
Collaborated in a 7-member team integrating MindsDB and Databricks for enterprise AI workflows. Developed APIs connecting Slack and enterprise systems via OAuth 2.0, optimized PostgreSQL schemas, and enhanced logging and monitoring to improve observability across services.
7-member team
MindsDB + Databricks
OAuth2 Slack integration
Django PostgreSQL MindsDB Databricks Slack API OAuth 2.0
Cardiovascular Age Prediction — Published Research
End-to-end ML pipeline for cardiovascular age prediction from single-lead ECG. Collected data from 42 subjects using ESP32 + AD8232, producing 6,131 segments with 13 features. Trained Random Forest (R²=0.99) and ResNet-18 transfer-learning models (R²=0.87). Co-authored (equal contribution) and published in Circuits, Systems, and Signal Processing (CSSP, Springer), 2025 — DOI 10.1007/s00034-025-03048-2.
42 subjects · 6,131 segments
R²=0.99 (Random Forest)
Published · CSSP (Springer) 2025
Python Keras scikit-learn PyTorch ESP32 IoT Research
Led a team of 4 building a QGroundControl-based drone swarm controller supporting 10+ UAVs with 50Hz telemetry. Enhanced custom Qt/QML ground control, optimized swarm control with multiprocessing, and built a high-throughput MQTT telemetry pipeline (designed for ~10K msg/sec) with multithreading. Built Gazebo SITL multi-drone simulation and custom mission-planning tools, with Dockerized CI/CD and Vagrant-isolated environments.
Led team of 4
10+ UAVs · 50Hz telemetry
MQTT pipeline (~10K msg/sec design)
C++ Qt Python MQTT ArduPilot PX4 Gazebo Docker Vagrant
Industrial Parts Cost-Estimation Backend
Built REST APIs and normalized PostgreSQL schemas for an industrial parts cost-estimation and quote-generation platform. Implemented data-ingestion pipelines transforming UI-submitted manufacturing data into structured backend records, and created authenticated access patterns so AI/ML services could fetch user-specific estimation data for automated quote generation.
Normalized PostgreSQL schemas
Authenticated data APIs
AI/ML quote-generation feeds
Python REST APIs PostgreSQL Data Ingestion AI/ML Integration
AI Prompt Management Platform
Full-stack Django/FastAPI platform integrating OpenAI GPT-4 with streaming responses for prompt versioning, with modules for prompt templates, function helpers, and task planning. Implemented RBAC access control and secure S3 asset storage, deployed via Docker and CI/CD. Built at VeritusLabs for internal AI workflow management.
Automated AWS infrastructure provisioning with Boto3, including VPCs, EC2, IAM, and ECS services. Deployed Flask apps on ECS with Docker and ensured public access via DDNS. Streamlined modular, repeatable deployments end to end.
VPC + EC2 + ECS automation
ECS Fargate deployment
DDNS public access
Python Boto3 AWS ECS Flask Docker Dynu DDNS
Distributed Job Scheduler
TCP/IP distributed scheduler in C with Pthread concurrency supporting 50+ simultaneous jobs. Implements process isolation via Fork/Exec for safe parallel execution.
50+ simultaneous jobs
Fork/Exec isolation
C TCP/IP Multithreading Distributed Systems
Real-Time Pub-Sub Messaging System
Publisher-subscriber messaging system with Redis in-memory caching and PostgreSQL persistence, with delivery acknowledgments. Built as a systems project demonstrating messaging, caching, and persistence design.
Redis + SQL persistence
Delivery acknowledgments
Python Redis PostgreSQL Distributed Systems
Personal Portfolio — mubashir-rehman.is-a.dev
Fully static personal portfolio built with Astro 5 — every route renders to HTML at build time from a single JSON data layer, with React reduced to small hydrated islands (an AI chatbot grounded in that same data, and the theme toggle). Features light/dark theming, a markdown-rendered journal, role-targeted landing pages, Astro view transitions, and a full SEO/AEO layer: JSON-LD structured data (Person, FAQPage, BreadcrumbList, BlogPosting, ScholarlyArticle), Open Graph, auto-generated sitemap, llms.txt, and AI-crawler-friendly robots.txt. Deployed to GitHub Pages via GitHub Actions CI/CD.
Marketing site for a rooftop chai café in Barkat Market, Lahore — designed, built, and deployed end to end. Astro 5 + Tailwind with a Swiper photo gallery, local-SEO metadata and sitemap, and geolocation-aware directions, shipped on Vercel. A small, real client-style engagement: from brief to live site.
Designed → deployed end to end
Fully static Astro, zero backend
Local SEO metadata + sitemap
Astro Tailwind CSS TypeScript Swiper Vercel Local SEO
Multi-level endless runner for Android in Unity/C# using object pooling to reduce memory usage. Implemented level progression, power-ups, and dynamic obstacles, and integrated Google Play Services for leaderboards and achievements. Achieved 5K+ downloads with a 4.2★ rating.
5K+ downloads
4.2★ rating
Object pooling
Unity C# Android Game Dev
Endless Runner Web Game
Web-based endless runner built with PlayCanvas and JavaScript. Optimized with scene pre-loading, object pooling, and reusable assets to minimize load times, with a scalable architecture for adding new levels.