Alex Ndungu

Co-Founder + CTO

Let's talk
HomeAboutExperienceProjectsSkillsContact
Let's talk
HomeAboutExperienceProjectsSkillsContact

Alex Ndungu

Backend systems, machine learning retrieval, and clean product-minded engineering for teams that care about reliability.

GitHubLinkedInLeetCodealexmeta517@gmail.com
Experience

A timeline shaped by backend rigor and domain-sensitive systems.

My experience sits at the intersection of production software engineering, structured data workflows, and domains where system reliability matters.

Apr 2026 — CurrentDeveloper platform / Kenya

CodePinioncodepinion.co.ke

Co-Founder & CEO

Co-founded CodePinion to replace the four-vendor toolchain software teams assemble by hand — git host, cloud IDE, PaaS, and project tracker — with one AI-native platform. I set the product direction and architected and built the entire stack; it runs in production on Kubernetes, serves 4 business clients plus 30+ independent developer signups, and is scaling toward becoming the default place African software teams build and ship.

  • - Set the company's technical strategy and product direction, then architected and built the full platform end to end — backend, frontend, infrastructure, and the AI execution layer.
  • - Built a pluggable runtime abstraction over Kubernetes and Docker execution backends, provisioning on-demand containerised dev workspaces with live terminals, public-URL dev-server previews, and language-server integration across seven languages — Python, TypeScript/JavaScript, Java, Go, Ruby, C#, and PHP.
  • - Designed an agentic tool-calling system on a tiered permission model that mirrors human dashboard authorization exactly — write actions route through a human confirmation round-trip, and irreversible actions additionally require typed confirmation plus an emailed OTP.
  • - Built the multi-provider AI routing layer beneath it: Anthropic and OpenAI support with bring-your-own-key credentials, per-organisation budget enforcement, and automatic model failover.
  • - Shipped a full git hosting layer — HTTP/SSH gateway, pull-request review, and CI/CD pipelines — with a stateless workspace push-token scheme that needs no database writes and survives workspace resume.
  • - Delivered infrastructure-as-a-product: managed databases, caching, message queues, search, and S3-compatible object storage provisioned as owner-scoped first-class resources alongside one-click PaaS app hosting.
  • - Designed the monetisation architecture around M-PESA rather than card-on-file billing, building for how the African market actually pays instead of retrofitting a Western payment assumption.
Django REST FrameworkNext.jsTypeScriptKubernetesDockerPostgreSQLCeleryRabbitMQMeilisearchAnthropic APIOpenAI APIWebSocketsFastAPI
Jun 2025 — Current (part-time)Proptech / Kenya-first SaaS

Harlem Manageharlem.co.ke

Co-Founder & CTO

As co-founder and CTO I own the technical direction of Harlem Manage, a Kenya-first operating system for property firms. I designed and built the platform — multi-tenant core, embedded AI agents, real-time M-PESA payment flows, async task processing, and a production observability stack — it is live with 3 property owners and 25+ unique logins across their teams, and I am scaling it as the system landlords, agencies, and property firms run their operations on.

  • - Own technical strategy and architecture as CTO, translating how Kenyan property firms actually operate into a platform they can run their business on rather than a generic management panel.
  • - Designed a role-based AI agent system with dynamic tool selection per user role, backed by full audit trails storing tool calls, results, and response payloads per message for compliance and debugging.
  • - Integrated Safaricom M-PESA C2B STK Push with OAuth2 token management and field-level encrypted credential storage per organisation and property; streamed live payment status to clients via JWT-authenticated WebSocket consumers over Redis channel layers.
  • - Designed a distributed async task architecture over Celery and RabbitMQ with publisher confirms and per-worker prefetch limiting, driving scheduled ledger generation, recurring maintenance automation, and payment reminders.
  • - Built a production observability stack integrating Sentry with GlitchTip — structured JSON logging, per-request UUID tracing, and a recursive PII scrubber that redacts sensitive field patterns before transmission.
  • - Modelled multi-tenancy so each organisation and property is fully isolated, with per-tenant payment credentials and financial settings supporting independent payment flows.
Django REST FrameworkPostgreSQLOpenAI APIDjango ChannelsWebSocketsCeleryRabbitMQRedisM-PesaDaphne/ASGISentryGlitchTip
Oct 2025 – Jul 2026Nairobi, Kenya

Savannah Informatics

Software Engineer

Worked across two production healthcare systems — a national health authority integration platform and the Slade360 payer product — with a focus on backend correctness, report delivery infrastructure, production debugging, and observability.

  • - Designed a shared link-generation service adopted across Slade360 after large report exports began failing silently against the email attachment size ceiling — moved every affected report from attached documents to generated download links, cutting report delivery time by roughly 80%.
  • - Designed the Payer Portal refresh-request workflow that pulls member data directly from insurance providers — introduced the refresh request model and refresh logic, extended it to the EDI pipeline, and built the Heritage adapter connection to e-oxygen to fetch specified member records, replacing manual data reconciliation.
  • - On the SHA (Social Health Authority) integration: captured preauthorisation expiry dates from the SHA adapter and persisted them to the preauth table, enabling downstream systems to correctly determine whether a provider could initiate a hospital visit — directly impacting patient access to treatment in Kenya.
  • - Diagnosed and resolved a production Celery worker stall caused by a MultipleItemsFound exception in an authorization task queue — traced to a duplicate-record conflict in an update-or-create lookup that caused the worker to crash silently.
  • - Onboarded and trained new engineers on Slade360 workflows, pairing on their tickets through to release; ran root-cause analysis across the Slade ticket backlog and reviewed release-reservation workflow RFCs.
  • - Repaired the broken CI/CD pipeline for a standalone service, restoring automated build and deploy for a system that had been shipping manually.
  • - Applied observability practices using Grafana, Sentry, Loki, and Prometheus — tooling I subsequently implemented in Harlem Manage's own production stack.
PythonDjangoCeleryCI/CDGrafanaSentryLokiPrometheusHealthcare data workflowsEDIREST APIs
3 monthsFintech / Nairobi, Kenya

E&M Tech

Software Engineer Intern / Contract

Worked across two production systems: a CBK-compliant banking data integration platform and a school management system, gaining hands-on experience in regulatory data workflows and structured backend design.

  • - Implemented complex Excel template upload for the Granular Data Integration (GDI) project — a CBK-mandated system that defined how banks and financial institutions submit structured data to the Central Bank of Kenya.
  • - Conducted data analysis on CBK templates to reverse-engineer their structure, which informed the upload implementation approach and ensured format compliance.
  • - Wrote tests that streamlined the GDI pipeline, catching edge cases in template parsing and improving reliability of the upload workflow.
  • - Implemented backend data structures for a school management system that powered structured data presentation to the frontend.
Excel template parsingData analysisBackend data structuresCBK complianceTestingFinancial data workflows