Distributed Systems at Scale
Event-driven, high-throughput systems that stay real-time under production load: partitioning, messaging, and stream processing with Kafka and Redis.
Software Architect
Martin Kostov
I design large-scale, real-time systems that stay fast under heavy traffic and honest about what they cost. I also put AI to work in engineering, and I test what it builds beyond the happy path.
Based in Sofia, Bulgaria. Conference speaker and published researcher.
What I work on
Performance & Reliability
Latency, throughput, and availability that hold when traffic spikes: caching, profiling, and resilience designed in before the incident, not after.
Cost-Aware Architecture
TCO, ROI, and infrastructure cost as first-class design inputs, so decisions hold up under real production and budget constraints.
AI, LLMs & Agents
LLM workflows and AI agents in engineering: guardrails, evaluation, and local LLMs alongside hosted models, backed by peer-reviewed research on LLM-assisted optimization.
Testing & Verification
Making AI-generated code prove itself: integration testing, performance testing, and human review beyond the happy path.
Writing & Books
The Offline Years, literary coming-of-age fiction set in Bulgaria before smartphones. Book One, The Circus Money, is in development.
Talks & workshops
Mastering the Total Cost of Ownership of Software
What software really costs: TCO, ROI, hidden engineering costs, risk, and how to put architecture trade-offs in terms stakeholders can act on.
Workshop detailsTesting AI-Generated Code Beyond the Happy Path
Where AI-generated code fails in production, and the integration, performance, and human checks that catch it before it ships.
Workshop detailsArticles & case studies
Betting History Data Propagation
How a high-traffic platform moves bet history at scale: raising throughput while cutting reporting latency, operational risk, and cloud cost.
Read on MediumSaving Costs While Improving KPIs with AMD EPYC
An infrastructure migration to AMD EPYC, validated with stress testing: Redis-backed performance held, KPIs improved, and the bill went down.
Read on MediumResearch
Feedback-Driven SQL Optimization with Validated Query Rewrites
Electronics (MDPI), 2026. Local LLMs propose SQL rewrites and the system trusts none of them: every candidate is empirically validated, so each accepted rewrite provably returns the same results, faster.
Read the paperBooks & fiction
The Offline Years
Literary coming-of-age fiction set in Bulgaria before smartphones: friendship, first love, and the years nobody photographed.
Book One, The Circus Money, is in development.
About
Martin Kostov is a software architect in Sofia, Bulgaria. He works on real-time systems designed for the worst minute of the day: high-traffic production where spikes are routine, events never stop arriving, and speed, reliability, and cost all have to hold at once.
He applies one rule to AI in engineering: it has to show its work. His peer-reviewed research in Electronics (MDPI) has local LLMs propose SQL rewrites, then validates each one empirically before it counts. His talks and workshops bring the same habit to the total cost of ownership of software and to testing AI-generated code beyond the happy path.
Away from production systems, he writes fiction. The Offline Years is a literary coming-of-age story set in Bulgaria before smartphones, when nothing was saved unless someone remembered it. Book One, The Circus Money, is in development.
Common questions
- Who is Martin Kostov?
- Martin Kostov is a software architect based in Sofia, Bulgaria. He designs large-scale, real-time systems that stay fast under heavy traffic and honest about what they cost, works on AI agents and LLM-based workflows in engineering, and has published peer-reviewed research on LLM-assisted SQL optimization.
- Has Martin Kostov published research?
- Yes. He is the author of Feedback-Driven SQL Optimization with Validated Query Rewrites, published in Electronics (MDPI) in 2026. In the paper, local LLMs generate SQL query rewrites and each rewrite is empirically validated before it is accepted.
- What is cost-aware architecture?
- Cost-aware architecture means designing and evolving systems with total cost of ownership in mind: balancing performance, reliability, infrastructure cost, and risk so that technical decisions hold up under real production and budget constraints.
- Does Martin Kostov speak at conferences or run workshops?
- Yes. He runs a full-day workshop, Mastering the Total Cost of Ownership of Software, and a talk and workshop, Testing AI-Generated Code Beyond the Happy Path. His full speaker profile is on Sessionize.
- What is The Offline Years?
- An adult literary coming-of-age fiction project about friendship, memory, and growing up in Bulgaria before smartphones. Book One, The Circus Money, is in development. Read more on the book site.
Get in touch
A system to scale, a workshop or talk invitation, or a question about the paper or the book? Send a message.