Open to opportunities

Zekarias Mesfin

Software Engineer

Crafting high-performance web apps and cloud infrastructure — based in Dubai, UAE.

View My WorkGet in Touch
Zekarias Mesfin profile
Dubai, UAE
6+
Years Experience
2×
AWS Certified
20+
Projects Delivered
Scroll
Who I Am

About Me

I am a Software Engineer based in Dubai with 6+ years of experience in full-stack web development, AI/LLM integration, cloud technologies, and systems programming. I specialize in building scalable, high-performance applications using modern frameworks such as Next.js, React, Django, Node.js, Go, and Rust — including production SaaS platforms with RAG pipelines, multi-tenant architecture, real-time WebSocket systems, and AI answer engines that stream cited, web-grounded responses.

Currently working as a Software Engineer, I bring a unique blend of technical expertise and business acumen. I hold dual AWS certifications (Solutions Architect Associate and Cloud Practitioner), reflecting my commitment to cloud-native development and DevOps best practices.

I am passionate about clean code, thoughtful architecture, and delivering solutions that create real impact — from cutting average API response times under peak load to running multi-channel AI messaging that answers customers in seconds rather than hours.

6+
Years Experience
2×
AWS Certified
Dubai
UAE Based
Full-Stack
Engineer
View Resume
What I've Built

Skills & Experience

1

Full-Stack E-commerce Solutions

Next.jsReactDjangoDRFPostgreSQLTailwind CSSNginxPayPalStripe
  • ▸Built scalable e-commerce platform serving 1,000+ products with full shopping cart, checkout, and payment processing
  • ▸Implemented PayPal and Stripe payment gateway integrations, processing $50K+ in transactions with PCI compliance
  • ▸Developed 15+ RESTful API endpoints for product management, order processing, and real-time inventory updates
  • ▸Deployed production applications on VPS with Nginx reverse proxy, SSL certificates, and 99.5% uptime
  • ▸Created seamless browsing experience with automatic invoice generation and digital delivery systems
2

Enterprise ERP & Business Systems

DjangoOdoo ERPPostgreSQLPythonREST APIsNext.js
  • ▸Developed custom ERP modules streamlining operations for 200+ users across inventory, sales, and accounting
  • ▸Engineered bi-directional Odoo integration for automated invoice and payment synchronization, eliminating manual data entry for 500+ transactions daily
  • ▸Created 10+ custom Odoo modules with external API integrations for extended functionality
  • ▸Built office management systems for room booking, IT help desk tickets, and hardware asset lifecycle tracking
  • ▸Automated payroll calculations, deductions, and reporting processes for organizational efficiency
3

Multi-Tenant SaaS & Omnichannel Platforms

DjangoDjango ChannelsDaphnedjango-tenantsRedisCeleryPostgreSQLWebSocketMeta APIsWhatsApp Cloud APITelegram Bot API
  • ▸Architected and launched Mels (mels.live) — a production multi-tenant SaaS omnichannel inbox aggregating WhatsApp, Instagram, Messenger, and Telegram into a unified real-time interface
  • ▸Implemented full PostgreSQL schema-per-tenant isolation via django-tenants, supporting unlimited tenant onboarding with automatic subdomain provisioning and on-demand HTTPS via Caddy
  • ▸Built real-time inbox with Django Channels + WebSocket, pushing new messages to all connected agents instantly without polling
  • ▸Engineered encrypted credential storage (AES-256-GCM) for all channel tokens and API keys at rest
  • ▸Designed subscription plan system with per-tenant token quota enforcement (daily + monthly caps) across multiple AI providers
  • ▸Serve seven business tenants across WhatsApp, Instagram, Messenger, and Telegram, with the AI assistant answering inbound customer messages at a median first response of 6 seconds
4

API Development & Integration

Django REST FrameworkExpress.jsNode.jsPostgreSQLMongoDBOAuthJWT
  • ▸Architected and deployed scalable APIs serving 50K+ requests daily across multiple applications
  • ▸Optimized database queries and API endpoints, reducing average response time from 800ms to <200ms
  • ▸Integrated third-party services including OAuth providers, payment gateways, and analytics platforms
  • ▸Built 15+ production-grade RESTful API endpoints handling authentication, authorization, and data management
  • ▸Implemented webhook integrations for real-time data synchronization with external systems
5

AI Integration & RAG Systems

Google Gemini APIGroq APIAnthropic Claude APIOpenAI APIpgvectorRAGLLMVector EmbeddingsNode.jsPython
  • ▸Built full RAG pipeline for Mels: documents (PDF, Excel, FAQ text) chunked and embedded via Gemini text-embedding-004 (768-dim vectors), stored in PostgreSQL with pgvector, retrieved via cosine similarity search at inference time
  • ▸Implemented multi-provider AI abstraction layer supporting Groq, Google Gemini, OpenAI, and Anthropic Claude — swappable per tenant with platform-managed encrypted API keys
  • ▸Integrated Gemini Vision (gemini-2.5-flash) for automatic image description of customer-sent media, injecting context into the AI system prompt for intelligent image-aware replies
  • ▸Built AI campaign generation pipeline: RAG-informed copy generation (3 ad concept variations) + Gemini image generation for Meta Ads publishing
  • ▸Developed database-driven AI prompt management CMS for Dante Trading — admins edit live Gemini system prompts, variables, output schemas, and switch models via UI with streaming SSE live preview
  • ▸Built automated AI blog content engine using Gemini 2.5 Flash: RSS feed ingestion, relevance scoring, quality-gated generation (≥800 words), Telegram approval workflow, and multi-channel publishing to Telegram + LinkedIn
  • ▸Tracked and enforced per-tenant LLM + Vision token quotas with atomic DB updates, preventing AI cost overruns across FREE/BASIC/BUSINESS/ENTERPRISE plan tiers
6

Systems Programming — Go & Rust

GoRustTokioAxumRayonSQLitegoroutinesRwLockClap CLI
  • ▸Built high-performance background job processing system in Go with zero external dependencies — goroutine worker pool (5 workers), buffered channel (200 cap), and thread-safe job store with real-time status tracking
  • ▸Developed production-ready URL shortener in Go with full click analytics (IP, User-Agent, Referer tracking), per-link stats, and SQLite persistence for zero-configuration deployment
  • ▸Engineered blazing-fast CLI log analyzer in Rust using Rayon parallel I/O, processing millions of log lines in seconds with regex keyword search and JSON/CSV export
  • ▸Implemented in-memory key-value store in Rust with Tokio + Axum HTTP API, TTL-based key expiry with async background reaper, concurrent read/write via RwLock, and JSON snapshot persistence
  • ▸Applied systems-level concurrency patterns: goroutines & channels (Go), async/await with Tokio, and data-parallel processing with Rayon (Rust)
7

Cloud & DevOps Infrastructure

AWS (S3, RDS, SNS, SQS, CloudFront)Dockerdocker-composeCI/CDGitHub ActionsLinuxNginxCaddyPM2Cloudflare R2
  • ▸Established CI/CD pipelines with GitHub Actions, automating testing, builds, and deployments across 15+ projects
  • ▸Deployed and managed production SaaS applications on AWS using S3, RDS, CloudFront, and other services
  • ▸Configured Caddy reverse proxy with on-demand TLS for Mels — new tenant subdomains get automatic HTTPS certificates with zero manual intervention
  • ▸Managed Cloudflare R2 object storage for tenant media files with per-tenant storage quota enforcement
  • ▸Configured containerized environments with Docker and docker-compose for consistent development and production deployments
  • ▸Set up Nginx reverse proxy, SSL certificates, and PM2 process management ensuring 99.5%+ uptime across multiple VPS deployments
8

AI Answer Engines & Conversational Products

FastAPINext.jsSSEGemini 2.5 FlashTavilyCohere Rerank v3.5PostgreSQLRedisSQLAlchemyAlembicAPSchedulerpython-telegram-botpytest
  • ▸Built Real Search (realsearch.tech) — a production AI answer engine that returns cited, web-grounded answers streamed token by token over Server-Sent Events
  • ▸Engineered a six-stage retrieval pipeline: query understanding and routing, Tavily web search, content extraction, Cohere Rerank v3.5 reranking, Gemini 2.5 Flash generation, and inline citation binding back to source URLs
  • ▸Split model usage by workload — Gemini 2.5 Flash for answer generation, Flash-Lite for routing, titles, and follow-up suggestions — cutting cost per query without degrading answer quality
  • ▸Hardened the fetch path against SSRF and added Redis-backed rate limiting plus search caching to keep third-party API spend bounded under load
  • ▸Persisted conversations, messages, and sources in PostgreSQL with SQLAlchemy and Alembic migrations, giving users resumable threads across sessions
  • ▸Built Focus Companion — a Telegram accountability coach running a morning check-in, silence-triggered smart nudges, and evening verdict loop, with streak tracking and weekly reports
  • ▸Implemented per-user timezone-aware scheduling with APScheduler so every user's morning, nudge, and evening jobs fire against their own local clock
  • ▸Designed a free/pro/elite tier system where paid tiers widen the AI coach's history window (7 → 14 → 30 days), letting it name multi-week patterns rather than react to a single day
  • ▸Persisted conversation state to SQLite so in-flight flows survive restarts and redeploys, and covered the product with a 15-module pytest suite that pins the published user guide to the bot's real behaviour
My Toolkit

Technical Skills

Frontend

React.jsNext.jsNext.js App RouterTypeScriptJavaScript (ES6+)Tailwind CSSHTML5CSS3ReduxContext APISSE Streaming UI

Backend

PythonDjangoDjango REST FrameworkFlaskFastAPIDjango ChannelsDaphne (ASGI)WebSocketsServer-Sent Events (SSE)Node.jsExpress.jsREST APIsGraphQLSQLAlchemyAlembicPydanticAsync PythonCeleryRabbitMQRedisAPSchedulerpython-telegram-botRate LimitingMulti-tenant Architecture

AI & LLM Integration

RAG PipelineVector EmbeddingspgvectorGoogle Gemini APIGemini 2.5 FlashGroq APIOpenAI APIAnthropic Claude APILLM IntegrationAI Prompt EngineeringGemini VisionAI Content GenerationAI Answer EnginesTavily Search APICohere RerankRetrieval & RerankingCitation GroundingQuery Understanding & RoutingLLM Token StreamingToken Quota Management

Systems Programming

GoRustTokioAxumRayonGoroutinesChannelsConcurrencyCLI ToolsParallel Processing

Databases

PostgreSQLMySQLMongoDBSQLitepgvectorPrisma ORMSQLAlchemy ORMSupabase (Postgres)Redis CachingDatabase Migrations (Alembic)Query OptimizationSchema Design

DevOps & Cloud

AWS (S3, RDS, SNS, SQS, CloudFront, CloudFormation)Dockerdocker-composeCI/CD (GitHub Actions)LinuxNginxCaddyGunicornUvicornPM2systemdCertbot / Let's EncryptVPS DeploymentCloudflare R2Turborepo

Tools & Practices

Git / GitHubGit FlowCode ReviewsJiraAgile / ScrumOAuthJWT AuthenticationWebhook Integrationspytestpytest-asyncioTest-Driven DevelopmentSSRF ProtectionTechnical DocumentationFramer MotionTelegram Bot API
What I've Made

Featured Projects

Mels — Omnichannel AI Inbox
View Project

Mels — Omnichannel AI Inbox

Production multi-tenant SaaS platform aggregating WhatsApp, Instagram, Messenger, and Telegram into a unified AI-powered inbox. Features a RAG chatbot with pgvector, embeddable web widget, Gemini Vision for image parsing, schema-per-tenant PostgreSQL isolation, real-time WebSocket updates, and AES-256-GCM encrypted credentials.

DjangoNext.jsRAGpgvectorMulti-tenantWebSocketGeminiSaaS
Visit Site
Real Search — AI Answer Engine
View Project

Real Search — AI Answer Engine

A production AI answer engine: ask a question, get a cited answer grounded in the live web, streamed token by token. Six-stage pipeline — query understanding and routing, Tavily web search, content extraction, Cohere Rerank v3.5, Gemini 2.5 Flash generation, and inline citation binding. Built on async FastAPI with SSE streaming, PostgreSQL-backed conversation history, Redis rate limiting and search caching, and SSRF-hardened URL fetching.

FastAPINext.jsGemini 2.5TavilyCohere RerankSSE StreamingPostgreSQLRedis
Visit Site
Focus Companion — AI Accountability Coach
View Project

Focus Companion — AI Accountability Coach

A Telegram bot that runs a full daily accountability loop: morning energy check-in and goal capture, smart nudges when you go quiet, and an evening done-or-missed verdict that builds a consecutive-day streak. An AI coach powered by Gemini 2.5 Flash reads up to 30 days of history to name patterns instead of platitudes. Features per-user timezone scheduling via APScheduler, free/pro/elite tiers with quota enforcement, persisted conversation state that survives restarts, admin tooling, and a 15-module pytest suite that pins the user guide to the bot's real behaviour.

PythonTelegram Bot APIGemini 2.5APSchedulerSQLiteDockerpytest
Open in Telegram
Dante Trading Digital Platform
View Project

Dante Trading Digital Platform

Full-stack digital platform for an industrial machinery company: Next.js website with product catalog, a custom headless CMS with markdown editor and AI prompt control, and an automated blog engine that ingests RSS feeds, generates content via Gemini 2.5 Flash, routes drafts through a Telegram approval bot, and cross-posts to LinkedIn.

Next.jsTurborepoPrismaGemini AITelegram BotCMSLinkedIn API
Visit Site
Imperfect Design Studio
View Project

Imperfect Design Studio

A 3D and interior design service platform with seamless browsing, PayPal payment integration, and automatic invoice generation. Currently in production use by the client.

Next.jsNode.jsExpress.jsPayPal
Visit Site
Go Task Queue
View Project

Go Task Queue

High-performance background job processing system built in Go with zero external dependencies. Features a goroutine worker pool, thread-safe job store, real-time status tracking (pending → running → completed/failed), and a clean REST API for submitting and monitoring jobs.

GoGoroutinesREST APIConcurrency
View on GitHub
Go URL Shortener
View Project

Go URL Shortener

Production-ready URL shortener with full click analytics. Tracks IP, User-Agent, and Referer on every redirect. Provides per-link stats including total clicks, last access time, and top referrers. Built with Go and SQLite for zero-configuration deployment.

GoSQLiteAnalyticsREST API
View on GitHub
Rust Log Analyzer
View Project

Rust Log Analyzer

Blazing-fast CLI log file analyzer built in Rust using parallel I/O via Rayon. Processes millions of log lines in seconds. Supports multi-file analysis, level filtering, regex keyword search, summary reports, and JSON/CSV export.

RustCLIRayonParallel Processing
View on GitHub
Rust KV Store
View Project

Rust KV Store

High-performance in-memory key-value store with an HTTP REST API, built with Rust + Tokio + Axum. Features TTL-based key expiry with async background reaper, concurrent read/write with RwLock, file persistence via JSON snapshots, and CORS support.

RustTokioAxumREST API
View on GitHub
Coffee Trade Platform
View Project

Coffee Trade Platform

Comprehensive Django platform facilitating international coffee trade between importers and exporters. Features real-time messaging, automated workflow management with phase progression, role-based access control, and comprehensive document handling for trade compliance.

DjangoBootstrapPostgreSQLReal-time
Visit Site
Cryptocurrency Price Telegram Bot
View Project

Cryptocurrency Price Telegram Bot

Real-time cryptocurrency price tracking Telegram bot built with FastAPI and python-telegram-bot. Features async architecture, RESTful API integration with CoinGecko, support for major cryptocurrencies (BTC, ETH, SOL, XRP), and comprehensive error handling.

PythonFastAPITelegram Bot
View on GitHub
Office Room Booking & IT Asset Management
View Project

Office Room Booking & IT Asset Management

SmartOffice web application for managing office room reservations, IT help desk tickets, and hardware asset inventory with lifecycle tracking.

DjangoPythonPostgreSQL
View on GitHub
Payroll Manager
View Project

Payroll Manager

Streamlined payroll management application that automates salary calculations, deductions, and reporting for organizations.

DjangoPython
View on GitHub
Madish
View Project

Madish

Restaurant order tracking application that simplifies placing and monitoring orders with a transparent process for customers and staff.

DjangoPython
View on GitHub
Razermind
View Project

Razermind

Marketplace for selling graphic design assets including logos, templates, and illustrations. Features a clean storefront, secure checkout, and instant digital delivery.

Web AppE-commerce
View on GitHub
More on GitHub
My Background

Education

Master of Business Administration (MBA)

Data Analytics & AI

Britts Imperial University College

Sharjah, UAE · March 2025 — Present

Research focus: AI model optimization, predictive analytics, and machine learning applications

Bachelor of Science

Computer Science & Engineering

Adama Science and Technology University

Adama, Ethiopia · September 2017 — July 2022

Credentials

Certifications

AWS Certified Solutions Architect — Associate

Amazon Web Services

Valid: Feb 2024 — Feb 2027

AWS Certified Cloud Practitioner

Amazon Web Services

Valid: Oct 2023 — Oct 2026