Skip to content

Sheet 01 / 04 · General arrangement

Yohannes Merga

Backend-focused full-stack engineerNode.js · TypeScript · NestJS · PostgreSQL · AWS

Drawn by
Y. Merga
Revision
v1.2 · 2026-07
Scale
4+ yrs : 1 page
Status
Open to backend roles

I default to Node and TypeScript, reach for Python where its ecosystem wins, and draw the diagram before I write the code, so the next engineer knows why it looks like this.

Four-plus years across an AI recruitment platform, a gaming payments stack, an ERP modernization and a bank's internal tooling. Solo backend owner at Mereb and frontend team lead on the same product. Express and NestJS services are my default shape; I use FastAPI and Celery for AI and data pipelines, and I have kept new services talking to a legacy PHP core without breaking it. Equally comfortable tracing a Postgres connection leak and shipping the React table that consumes the endpoint.

Experience
4+ years across 6 teams
Signups
10× after the ATS rebuild
Screening
faster recruiter selection
Search
80% faster candidate queries
Coverage
85% backend tests, gated in CI
Remote for
USA · Kuwait · Slovenia
FIG. 1

AI-powered ATS, service arrangement

Mereb Technologies · 2024 – 2025
FIG. 1: AI-powered ATS, service arrangementRequest path solid, background pipeline dashed. Express fronts the platform, Python workers carry the AI pipeline, and Redis brokers every async hop.CLIENT / EDGESERVICES · AWS ECSSTATERecruiter consoleReact 18 · TanStackCandidate portalNext.js · ZustandCloudFront + S3static assetsAPI gateway / BFFExpress.js · auth · rate limit · OpenAPI 3nodecandidatesingestion · profileasync SQLAlchemypyrecruitingjobs · pipelinesscoring rulespyAI pipeline workersLangChain → OpenAI · parse, summarize, extractstructured output → embeddings → scorepysandbox schedulerpre-provisions isolated IDE environmentsnodeRedis · broker · cache · locks · idempotencyPostgreSQLnormalized coreindexes · migrationsTypesensecandidate + job searchresync from PostgresS3 · raw CVspresigned uploadObservabilityOpenTelemetry spansPrometheus metricsstructured JSON logs[1] query patterns andindexes tuned for load[2] search kept besidePostgres, not inside it[3] every span carriescandidate_id + trace_idsync requestasync / queued
FIG. 1 · Request path solid, background pipeline dashed. Express fronts the platform, Python workers carry the AI pipeline, and Redis brokers every async hop.Scroll the drawing sideways to pan.
FIG. 2

Payments and observability over a legacy core

Real Gaming · 2025 – 2026
FIG. 2: Payments and observability over a legacy coreNew services never write to the legacy core directly. REST and gRPC facades hold the contract, RabbitMQ carries every side effect, and one provider interface hides three PSPs.CLIENTSSERVICE LAYEREXTERNAL / LEGACYGame clientin-game purchasesWeb appmarketplaceAdmin toolsops and supportREST + gRPC facadeversioned contracts · one shape per consumerpymarketplacelistings · orderspymembershiptiers · rewardspywageringin-game transactionspypaymentsone provider interfacepyRabbitMQ · topic exchange · durable queuesevent consumersonboarding · rewards · wagering payoutsretries with backoffpylegacy coreexisting product surfacereached via facade onlyphpBraintreeStripeHyperwalletPostgreSQLpooled, tuned accessObservabilityOTel → PrometheusClickHouse · HyperDXELK logs[1] gRPC keeps latencylow across the boundary[2] pool exhaustiontraced, pooling retuned[3] payout events areidempotent on retrysync requestasync / queued
FIG. 2 · New services never write to the legacy core directly. REST and gRPC facades hold the contract, RabbitMQ carries every side effect, and one provider interface hides three PSPs.Scroll the drawing sideways to pan.
TAB. 1

Measured outcomes

every number, and what it was measured against
Candidate registrations
10×
daily signups
Mereb, ATS platform
Recruiter selection speed
reduced screening time
Mereb, ATS platform
Search performance
+80%
query latency
Mereb, Typesense + PostgreSQL
Initial interview time
−50%
recruiter feedback
Mereb, AI pre-screening
Active users
+30%
monthly growth metrics
Real Gaming, marketplace
API response time
−8%
latency benchmarks
Sulala, schema and index work
Data retrieval
−11%
query execution time
Sulala, query tuning
Issue resolution
response time
Siinqee Bank, internal tooling
Backend test coverage
85%
CI coverage reports
Sulala, Pytest suite
FIG. 3

Build log

reverse chronological · 6 entries
  1. 2024 – 202517 months · Addis Ababa

    Lead Software Engineer

    Mereb Technologiessolo backend owner, frontend team lead

    Architected and shipped a production AI-powered ATS end to end. Express.js services front the platform, async FastAPI services and Python workers carry candidate ingestion, AI processing and scoring, and Typesense sits beside PostgreSQL for search. Built the React 18 recruiter console on top, plus the Docker environments, GitHub Actions pipelines and AWS ECS deploys underneath.

    • 10× candidate registrations, measured by daily signups
    • 4× faster recruiter selection, measured by reduced screening time
    • 50% less initial interview time, measured by recruiter feedback
    • 80% faster search, measured by query latency
    • Provisioning wait eliminated, measured by near-zero wait to start a challenge
    • Express.js
    • TypeScript
    • FastAPI
    • LangChain
    • OpenAI API
    • PostgreSQL
    • Redis
    • Typesense
    • React 18
    • TanStack
    • AWS ECS/ECR
    • Pytest
  2. 2025 – 202611 months · remote, USA

    Backend Engineer

    Real Gaming

    Marketplace, membership, payment and in-game transaction services. Integrated Braintree, Stripe and Hyperwallet behind one payment interface, exposed REST and gRPC so new services could work with a legacy PHP core, and drove RabbitMQ event flows for onboarding, rewards and wagering. Added OpenTelemetry tracing with Prometheus, ClickHouse, HyperDX and ELK behind it.

    • 30% more active users, measured by monthly growth metrics
    • Recurring API outages resolved, measured by reduced downtime
    • Faster incident detection, measured by time to detect in ELK
    • Low-latency service calls, measured by request round-trip time over gRPC
    • gRPC
    • RabbitMQ
    • FastAPI
    • Braintree
    • Stripe
    • Hyperwallet
    • OpenTelemetry
    • Prometheus
    • ClickHouse
    • legacy PHP core
  3. 20254 months · remote, Slovenia

    Backend Engineer

    PRO-BIT Programska Oprema d.o.o.ERP system modernization

    Distributed service layer for an ERP modernization: a centralized service registry, TTL-based heartbeats, automatic deregistration of unhealthy instances, and health plus metadata endpoints feeding operational dashboards.

    • Fewer service discovery failures, measured by production incidents
    • Higher availability, measured by uptime
    • Real-time health visibility, measured by status endpoints on dashboards
    • Service registry
    • Health checks
    • Load balancing
    • Microservices
    • Docker
    • CI/CD
  4. 2023 – 20248 months · remote, Kuwait

    Backend Engineer

    Sulala

    RESTful services with Celery and Redis background processing. Reworked PostgreSQL schemas, indexes and queries to cut latency, added caching to remove redundant database work, and built the automated test suite that gates CI.

    • 8% faster API responses, measured by latency benchmarks
    • 11% faster data retrieval, measured by query execution time
    • 85% test coverage, measured by CI coverage reports
    • Celery
    • Redis
    • PostgreSQL
    • FastAPI
    • Pytest
    • Docker
  5. 2023 – 20248 months · Addis Ababa

    Full-Stack Engineer

    Siinqee Bank

    Internal full-stack tools and automation for banking operations: React dashboards over backend APIs, messaging bots for operational teams, Ansible and Docker provisioning, and third-party integrations for location verification and branch data.

    • 3× faster issue resolution, measured by response time
    • More consistent infrastructure, measured by deployment stability
    • Fewer customer complaints, measured by support tickets
    • Node.js
    • React
    • Ansible
    • Python
    • Docker
    • SQL
  6. 2022 – 20234 months · Addis Ababa

    Junior Full-Stack Developer

    EagleLion System Technology

    Business and financial workflow applications in React and TypeScript, Docker dev and deploy environments, GitHub Actions pipelines, and role-based BI dashboards.

    • More reliable deployments, measured by release success rate
    • Better reporting accuracy, measured by stakeholder BI dashboards
    • Stronger API quality, measured by test coverage in CI
    • TypeScript
    • React
    • REST
    • GitHub Actions
    • Docker
TAB. 2

Specification

underlined = daily driver
Node & TypeScriptdefault choice for new services
Node.jsTypeScriptExpress.jsNestJSPrismaJestVitest
Years: 4+
PythonAI pipelines, workers, data-heavy work
FastAPIDjango / DRFCelerySQLAlchemyPytest
Years: 4+
PHPkept a legacy core working while new services landed
legacy core integrationREST and gRPC facades over PHP
Years: 1
API & architecture
RESTOpenAPI / SwaggergRPCmicroservicesevent-drivenservice discovery
Years: 4+
AI & LLM
LangChainOpenAI APIembeddingsRAGdocument extractionprompt design
Years: 2
Data & search
PostgreSQLRedisTypesenseMongoDBMySQLClickHouseOpenSearch
Years: 4+
Messaging
RabbitMQKafkaCelerybackground workerscaching
Years: 3
Cloud & DevOps
DockerGitHub ActionsAWS ECS / ECR / EC2 / S3 / CloudFrontAnsibleGCPHetzner
Years: 4
Observability
OpenTelemetryPrometheusHyperDXClickHouseELKdistributed tracing
Years: 2
Payments
BraintreeStripeHyperwalletvaulted methodsservice-to-service payouts
Years: 1
Frontend
React 18Next.jsZustandTanStack QueryTanStack TableTailwind
Years: 4
Also shipped
Gon8n automationSolidityERC-20 / 721HardhatPolygon
Years: past