Alireza Rezapour
Backend Engineer
About Me
Backend engineer with 3+ years owning services from monolith-split to event-driven scale, plus production LLM/RAG work. Comfortable across Node.js, Go, Python and the messy edges in between (PostgreSQL tuning, Kong plugins, hybrid retrieval). Shipped wins: 90% fewer failed transactions, 99% read-load reduction, 70% MTTR cut, hallucination under 3% on a UK-lender RAG product.
Work Experience
Aug 2025 – Present
Backend Engineer • Mashhad, Iran
- Refactored payment-and-transfer engine core: tightened transaction isolation, optimized hot-path queries with targeted indexes, and tuned PgBouncer pooling — cutting transaction failures 90% and production incidents 30%.
- Introduced read-side caching on hottest endpoints, dropping DB read load 99% (from 5,000+ req/s to under 20 req/s).
- Migrated synchronous inter-service calls to RabbitMQ-based event-driven messaging, cutting cross-service network traffic 40% and decoupling failure domains.
- Migrated analytical workloads from PostgreSQL to ClickHouse, halving dashboard p95 latency and offloading ~50% of PostgreSQL load.
- Extended Kong API Gateway with custom Go plugins for auth, rate-limiting, and request shaping across production traffic.
- Drove team-wide AI-assisted development practices (Cursor .rules, Claude CLAUDE.md, prompt patterns), cutting dev cycle time ~20% and lowering LLM token spend through better tool and agent usage.
- Took ownership of legacy "What to Buy" top-4 crypto recommendation service: replaced underlying prediction LLM, added validation layer to filter low-confidence picks, improving recommendation quality.
Criterio (Sub-brand) — Lead Engineer
- Built Criterio, an agentic RAG assistant over 50+ UK-lender lending-criteria documents with multi-turn context memory, follow-up question generation, and citation-grounded answers; replaced pure-embedding retrieval with AI-labeled chunk lookup over SQLite to fix recall on small corpora, holding LLM hallucination rate under 3% via strict prompting and citation grounding, with sub-60s response latency.
- Shipped fully AI-driven daily blog auto-publishing one lending-criteria news article per day with generated imagery and cited sources. Stack: LangGraph, LangChain, Python, FastAPI, OpenAI / OpenRouter.
Apr 2024 – Aug 2025
Backend Engineer • Mashhad, Iran
- Resolved B2B-service outages (payments, auth, cron) via cross-team response and post-mortem ownership to prevent recurrence.
- Migrated ~12 backend services from pm2-based deploys to a new internal Kubernetes-like platform, cutting deploy time from ~1 working day to minutes; onboarded team to the new workflow.
- Designed and built B2B infrastructure services (payment, user, auth, cron) using hexagonal architecture and event-driven patterns to keep core domain logic decoupled and testable.
- Diagnosed peak-time socket exhaustion across services, traced root cause to synchronous HTTP fan-out from cron jobs; authored proposal for message-queue architecture later adopted org-wide.
- Authored RFCs on logging/observability, error-handling and API contracts, and service boundaries; standards adopted across multiple teams to reduce integration friction.
- Acted as primary incident point of contact, leading critical-outage calls and cross-team meetings on behalf of tech lead.
- Mentored junior engineers via code reviews, pair programming, and architectural guidance, raising team velocity and code-review quality.
Aug 2023 – Apr 2024
Backend Developer • Mashhad, Iran
- Contributed to monolith-to-microservices migration across core services, splitting domains and shortening deploy cycles.
- Redesigned ~200-symbol crypto market-data ingestion service for higher fetch throughput, unblocking real-time AI analytics.
- Adopted NATS for low-latency inter-service streaming, feeding downstream AI/analytics consumers.
- Deployed KrakenD API Gateway for routing, rate-limiting, and horizontal scaling of backend endpoints.
Jan 2023 – Aug 2023
Backend Developer • Mashhad, Iran
- Built Node.js/Express REST endpoints with MongoDB persistence on backend team.
- Implemented real-time WebSocket price-broadcast feature for live trading data.
- Integrated BEP20/ERC20 wallet flow: per-user wallets on signup, on-chain balance polling via cron, deposit reconciliation to DB.
Projects
AI assistant for UK mortgage brokers: natural-language Q&A over 50+ lender criteria docs with multi-turn memory and cited answers. AI-labeled chunk retrieval, <3% hallucination, sub-60s latency. Daily auto-published lending-news blog with generated imagery.
Persian-language fintech: automated trading bot for global markets plus self-custody crypto wallet with fee-free internal transfers. Android app on Café Bazaar, Myket, APK.
Other Projects
Storage engine for Multer enabling file uploads to MinIO; published to npm with external forks.
Self-built reverse proxy with Node cluster and load balancing.
Decentralized email PoC: encrypted messages between Ethereum addresses via IPFS and on-chain events.
Tic-tac-toe with LLM-powered trash-talking opponent: Vue UI + FastAPI backend (minimax + LLM banter).
Open Source
Skills
Languages
Javascript
Typescript
Node.js
GolangPythonFrameworks
Nest.jsExpress.js
Adonis.jsFastAPIDatabases
PostgreSQL
MongoDB
Redis
ClickHouse
InfluxDBMessaging
RabbitMQ
NATS
KafkaCloud & DevOps
DockerPodman
Kong
KrakenD
CI/CD (GitLab/GitHub)NginxS3Monitoring & Debugging
Grafana
OpenTelemetry
OpenSearch
Graylog
SigNozArchitecture & Practices
AI / LLM
Methodologies








