Skip to content
Rashad Musayev

Rashad MusayevSoftware Engineer, Backend & AI Systems

Budapest, Hungary

Engineering backend systems.Building practical AI.

I'm Rashad Musayev, a software engineer focused on scalable backend architecture, production-grade APIs and AI-powered systems. I build software where solid engineering and applied artificial intelligence meet.

Explore my workGet in touchDownload CV(PDF)
Illustrative backend and AI systemA client calls an API. The API writes to a PostgreSQL store and queues work in Redis for Celery workers. Workers call a language model, which also receives context retrieved with pgvector from the store.Clientweb, mobileAPIDjango RESTQueueRedisWorkersCeleryStorePostgreSQLRetrievalpgvectorModelLLM
Fig. 0Illustration.The general shape of the systems I build: an API, background workers, a relational store and an AI layer with retrieval. Not a specific deployment.
Experience
About four years building backend systems and production web platforms
Currently
Contract engineer at Speakday and freelance engineer for MECC Group
Academic
M.Sc. Computer Science, ELTE (2026); Ph.D. in Informatics in progress

About

Budapest, Hungary

Backend architecture, distributed systems and applied AI, across the whole lifecycle.

I work where backend engineering meets artificial intelligence. Most of my time goes into the parts users never see: the data model, the API contract, the background job that has to run exactly once, the prompt that needs the right context to be useful.

Over about four years I have built Django and PostgreSQL backends for tutoring, driver-education, e-commerce and language-learning platforms, microservices for a London-based company, and four production web platforms for a construction group. Alongside that work I completed an M.Sc. in Computer Science at Eötvös Loránd University in July 2026, with a thesis on multi-agent LLM systems for requirements engineering.

I have since been admitted to the Ph.D. Programme in Informatics at Széchenyi István University, where I am researching knowledge graphs and retrieval-augmented generation for IT operations.

Where I work in the lifecycle

  1. DesignBackend architecture, data models and service boundaries.
  2. BuildRESTful APIs and data-intensive applications.
  3. IntegrateLLMs and retrieval pipelines aimed at specific problems.
  4. HardenPerformance, security and reliability work on live systems.
  5. AutomateDeployment pipelines and event-driven operational workflows.

Expertise

Seven areas

The tools I reach for, grouped by the problems they solve.

The first three groups are the core of my work. The rest support it: they are how products get shipped, secured and kept running.

Backend engineering

The core of the work: APIs and services in Python, structured so they stay maintainable as they grow.

  • Python
  • Django
  • Django REST Framework
  • FastAPI
  • Flask
  • REST APIs
  • Microservices
  • Clean Architecture
  • Domain-Driven Design
  • Event-driven architecture

Applied in Speakday, Clisso, Codelandia, M.Sc. thesis

Data & distributed systems

Schemas, queries and indexes tuned for real load, with caching and background work where they pay off.

  • PostgreSQL
  • MySQL
  • MongoDB
  • Redis
  • pgvector
  • Celery
  • Database optimization
  • Query and index optimization

Applied in Clisso, Codelandia, Speakday, M.Sc. thesis

AI engineering

LLMs used for specific jobs, grounded in retrieved context and coordinated across agents.

  • Large Language Models
  • Retrieval-Augmented Generation
  • Multi-agent systems
  • Knowledge graphs
  • Prompt engineering
  • OpenAI API
  • AI-assisted software engineering

Applied in M.Sc. thesis, Speakday, Ph.D. research

Frontend & full-stack

Enough frontend to ship complete products end to end.

  • TypeScript
  • React
  • TanStack Start
  • Svelte
  • Tailwind CSS

Applied in MECC Group, Codelandia

Infrastructure & DevOps

Containerised services and automated delivery pipelines.

  • Docker
  • Kubernetes
  • GitHub Actions
  • GitLab CI/CD
  • Cloudflare Workers
  • AWS
  • Linux
  • Nginx
  • Gunicorn

Applied in MECC Group, Clisso, Codelandia

Security & quality

Controls built into the application, verified by tests.

  • Row-level security
  • Content Security Policy
  • Cloudflare Turnstile
  • Playwright
  • Automated testing
  • Secure API design
  • Code review

Applied in MECC Group, Speakday, Clisso

Real-time & audio

Live conversations and transcription feeding AI analysis.

  • LiveKit
  • Speechmatics
  • Real-time conversation infrastructure
  • Speech-to-text integrations

Applied in Speakday

Experience

2023 to present

Production systems, from tutoring platforms to real-time AI.

  1. Jul 2026 to Present

    Remote

    SpeakdaySoftware Engineer (Contract)

    Backend engineering for an AI-powered spoken-language practice platform: live conversations, speech-to-text and automated feedback for learners. Changes ship through a spec, implement, review and deploy workflow across UAT and production.

    • Replaced a cron-based session-analysis trigger with an event-driven one, so AI feedback reports are generated as soon as a conversation ends rather than at the next scheduled run.
    • Fixed per-participant audio trimming and transcript sentence derivation, then wrote repair and backfill management commands that run as a dry run first and apply only when the output checks out.
    • Reworked the LLM prompts behind key insights, vocabulary, practice exercises and weekly conversation generation, and exposed the results through the API.
    • Closed a data-exposure issue on a public API endpoint and added per-channel notification preferences.
    • Redesigned the admin session-report view and removed a legacy Janus media-server dependency.
    • Python
    • Django
    • Celery
    • PostgreSQL
    • LiveKit
    • Speechmatics
    • LLMs
    See the feedback pipeline

  2. May 2026 to Present

    Budapest, Hungary (Remote)

    MECC GroupFreelance Software Engineer

    Built and shipped four production web platforms for an international construction company and its residential brands: mecc.group, m21apartments.com, the Campus Residences marketing site and a shared internal admin platform.

    • Bilingual English/Hungarian marketing sites, server-rendered with TanStack Start on Cloudflare Workers, including apartment and floor-plan browsers.
    • Inquiry forms backed by Supabase and protected with Cloudflare Turnstile and rate limiting.
    • A central admin platform with single sign-on across three Supabase schemas, per-brand role-based access and content management for every site.
    • Hardened all four applications with CSP nonces, security headers, sanitized rich-text rendering and row-level security policies.
    • Playwright end-to-end tests and automated Cloudflare deployments through CI/CD.
    • Bilingual SEO blog publishing, plus an AI-assisted marketing content workflow built on a brand-voice knowledge base.
    • React
    • TypeScript
    • TanStack Start
    • Supabase
    • Cloudflare Workers
    • Playwright
    See how the four platforms connect

  3. Jan 2025 to Oct 2025

    London, United Kingdom (Remote)

    ClissoSoftware Engineer

    Full-time backend engineering on microservices-based applications built with Python, Django REST Framework and PostgreSQL.

    • Designed and optimized database schemas for high-volume data, tuning queries and indexes to bring latency down.
    • Worked with DevOps to set up CI/CD pipelines and streamline deployments, shortening release cycles.
    • Took part in Scrum planning and code review, and mentored other engineers.
    • Python
    • Django REST Framework
    • PostgreSQL
    • Microservices
    • CI/CD

  4. Feb 2023 to Sep 2024

    Baku, Azerbaijan

    Codelandia Software CenterSoftware Engineer

    Built a data-intensive tutoring platform with a Django REST Framework backend and a Svelte frontend.

    • Automated deployments with GitHub Actions and Docker, making releases more reliable.
    • Optimized API serialization and introduced Redis caching to cut response times.
    • Django REST Framework
    • Svelte
    • Redis
    • Docker
    • GitHub Actions

Projects

Selected work

Systems I have designed, built and shipped.

The diagrams are drawn for this page from what I can share publicly. They show how each system is shaped, not internal code or data.

Research meets applied AI engineering

AI Stakeholder Agents for Automated Requirements Engineering

M.Sc. thesis, Eötvös Loránd University

  1. 1

    Project documentation

    The source material each analysis is grounded in.

  2. 2

    Retrieval & context

    Relevant passages found through pgvector embeddings.

  3. 3

    AI stakeholder agents

    LLM agents carry out the stakeholder analysis.

  4. 4

    Ambiguity detection

    Unclear requirements are identified.

  5. 5

    Refined requirements

    Refinement supported with the findings.

Fig. 1Conceptual diagram.Conceptual flow of the thesis platform, drawn from its description. The stages are logical steps, not separately deployed services.

The problem

Requirements arrive incomplete and ambiguous, and checking them against every stakeholder’s point of view is slow manual work that usually happens too late.

My contribution

  • Designed a multi-agent platform in which LLM agents represent stakeholders and analyse requirements from their perspective.
  • Grounded the agents in the project’s own documentation with retrieval-augmented generation over pgvector embeddings.
  • Built the backend that detects ambiguities in requirements and supports refining them.

Technical approach

Django and PostgreSQL hold projects, documents and requirements; pgvector stores embeddings for retrieval; Celery and Redis handle background processing; the OpenAI API powers the agents.

  • Python
  • Django
  • PostgreSQL
  • pgvector
  • Redis
  • Celery
  • OpenAI API
  • RAG
  • Multi-agent systems

Contract backend engineering

Speakday

Real-time AI conversation platform

  1. 1

    Live conversation

    Real-time audio sessions over LiveKit.

  2. 2

    Audio & transcription

    Per-participant audio, transcribed by Speechmatics.

  3. 3

    Session completion

    Ending a session now triggers analysis directly.

    Trigger changed from a cron schedule to an event

  4. 4

    AI analysis

    LLM prompts for insights, vocabulary and exercises.

  5. 5

    Learner feedback

    A report delivered through the API.

Fig. 2Conceptual diagram.Conceptual feedback pipeline. No internal code, data or screenshots are shown.

The problem

Learners practise speaking in live conversations and expect useful feedback while the session is still fresh. Feedback that waits for a scheduled batch arrives late.

My contribution

  • Moved session analysis from a cron schedule to an event-driven trigger that fires when a conversation ends.
  • Fixed participant-level audio trimming and transcript sentence derivation, with dry-run-first repair commands for affected records.
  • Improved the prompts that generate insights, vocabulary and practice exercises.

Technical approach

LiveKit carries the live conversation, Speechmatics transcribes the audio, and the Django, Celery and PostgreSQL backend runs the LLM analysis that becomes learner feedback.

  • Python
  • Django
  • PostgreSQL
  • Celery
  • LiveKit
  • Speechmatics
  • LLMs

Freelance, design to deployment

MECC Group digital ecosystem

Four production platforms for a construction group

The problem

A construction group and its residential brands needed bilingual marketing sites, apartment discovery and lead collection, without each brand running its own admin.

My contribution

  • Three server-rendered EN/HU marketing sites on Cloudflare Workers, with apartment and floor-plan browsing.
  • One admin platform with single sign-on and per-brand roles that manages content and leads for every site.
  • Security and delivery across all four: CSP nonces, row-level security, Turnstile, Playwright tests and CI/CD deployments.

Technical approach

React and TanStack Start render each site at the edge. Supabase provides auth and a schema per brand, and the shared admin reads and writes all of them under row-level security.

Three bilingual marketing sites, mecc.group, m21apartments.com and Campus Residences, are server-rendered on Cloudflare Workers. Each reads its content from and sends inquiries to its own schema in Supabase. One shared admin platform signs staff in once and manages content and leads for all three brands.

Fig. 3Conceptual diagram.How the four platforms relate. Simplified; infrastructure details are omitted.
  • React
  • TypeScript
  • TanStack Start
  • Supabase
  • Cloudflare Workers
  • Playwright
  • CI/CD
DDA stack sketchA Django REST Framework API backed by PostgreSQL, with Redis as a broker for Celery workers, all running in Docker.DockerAPIDjango RESTDatabasePostgreSQLBrokerRedisWorkersCelery
Fig. 4Stack sketch.Components as listed in the CV; the arrangement is illustrative.

DDA, Digital Driving Academy

Backend for a scalable driver education platform.

API design, data handling and background processing, packaged with Docker for repeatable deployments.

  • Django REST Framework
  • PostgreSQL
  • Redis
  • Celery
  • Docker
dda.az (DDA, Digital Driving Academy) (opens in a new tab)
LeadTech stack sketchStore clients call a Django REST Framework API that uses Redis and PostgreSQL, running in Docker.DockerClientrequestsAPIDjango RESTCacheRedisDatabasePostgreSQL
Fig. 5Stack sketch.Components as listed in the CV; the arrangement is illustrative.

LeadTech

API for an online electronics and home-appliance store, built for low-latency queries.

A performance-first e-commerce backend designed around low-latency queries, with PostgreSQL for storage and Redis alongside it.

  • Django REST Framework
  • PostgreSQL
  • Redis
  • Docker
leadtech.az (LeadTech) (opens in a new tab)

Other client projects

Research

Ph.D. in progress

Research that feeds back into engineering.

My Ph.D. work looks at how knowledge graphs and retrieval-augmented generation can support IT operations. It sits next to my engineering work rather than replacing it.
  1. 2026 – presentIn progress

    Ph.D. in Informatics

    Széchenyi István University, Győr, Hungary

    • Research focus: knowledge graphs and retrieval-augmented generation for IT operations.
    • A systematic literature review of the research domain is in preparation.
  2. 2024 – 2026Graduated July 2026

    M.Sc. in Computer Science

    Eötvös Loránd University (ELTE), Budapest, Hungary

    • Thesis: AI Stakeholder Agents for Automated Requirements Engineering.
    • Built the thesis backend with Python, Django, PostgreSQL, pgvector, Redis, Celery and the OpenAI API.
  3. 2020 – 2024Completed

    B.Sc. in Computer Engineering

    Azerbaijan State Oil and Industry University (ASOIU), Baku, Azerbaijan

Research focus concept mapKnowledge graphs, retrieval-augmented generation and large language models are connected to each other and to IT operations, the application domain of the research.Retrieval (RAG)KnowledgegraphsLLMsIT operations
Fig. 6Concept map.How my research interests relate. A representation of interests, not of published results.

Broader research interests

  • Artificial intelligence
  • Software engineering
  • Large language models
  • Knowledge graphs
  • Multi-agent systems

Principles

How I work

How I approach engineering.

Working habits rather than slogans. Each one points to where it has shown up in my work.
Reliability by design

Systems should be safe to operate, not only correct on the happy path.

Data repairs at Speakday run as a dry run first and are applied only once the output is verified.

Security as a requirement

Authorization, input handling and browser controls belong in the application from the start.

Row-level security, CSP nonces and sanitized rich text across the MECC platforms; a closed data exposure on a public API.

Performance through engineering

Speed comes from data access that has been measured and designed, not from guesswork.

Query and index tuning at Clisso; serialization work and Redis caching at Codelandia.

Practical AI integration

LLMs earn their place by solving a specific problem with the right context.

Retrieval over project documentation in the thesis; targeted feedback prompts at Speakday.

Automation over repetition

If a step happens on every release or every event, it should run on its own.

Event-driven analysis at Speakday; CI/CD deployments for MECC, Clisso and Codelandia.

Contact

Have a challenging system to build?

I'm interested in backend engineering, AI-powered applications and technically demanding products. If you're building something ambitious, let's talk.

rashadmusayev19@gmail.com
Location
Budapest, Hungary