Daniel Malungu

Dec 2021 – Present · Nairobi, Kenya

• Led ArdhiSasa through a national audit by the Office of the Auditor-General covering documentation, user guides, development workflow and methodology, architecture and implementation, and security practices — receiving a positive assessment in the Auditor-General's report on the State Department for Lands and Physical Planning for the year ended 30 June 2025
• Designed and administer PostgreSQL 17 production clusters with hybrid master-replica topology using repmgr HA and PgBouncer connection pooling
• Scaled cluster throughput from 1,000 TPS to 10,000 TPS through architectural redesign and query optimization
• Implemented SCRAM-SHA-256 password encryption across PostgreSQL clusters so the database never handles raw passwords during user creation or rotation, paired with audit triggers and automated backup scripts across 31+ databases
• Implemented S3-based long-term archival of database backups, VM security logs, application logs, and database logs to meet Kenya's statutory data retention requirements
• Defended the platform against an active attack by enforcing end-to-end TLS across every service hop — client-to-server, database connections, and support tooling such as Mattermost and pgAdmin — and restructured credential access around HashiCorp Vault for system-critical authentication
• Built and maintain ETL/ELT pipelines in Python, Apache Airflow, and SQL for land transaction and records data
• Implemented PeerDB change data capture replication from production OLTP to data warehouse with custom schema mapping
• Architected the data warehouse on Data Vault 2.0 and grew data engineering from a non-existent BI function into a fully independent team delivering all of ArdhiSasa's BI, visualization, and reporting on system-generated data for decision makers
• Led digitization of 5 million+ land records through a custom EDMS including a bulk document ingestion ETL tool
• Designed and implemented CI/CD pipelines with GitHub Actions for automated data pipeline testing and production deployments; led the migration of the deployment platform from Docker Swarm to a fully self-hosted, on-premises Kubernetes cluster to improve service delivery
• Improved database observability by rolling out pghero, pgBadger, and pgwatch for query performance monitoring, log analysis, and error tracing across production clusters
• Served as Scrum Master for the data engineering team — sprint planning, backlog grooming, and agile coaching

• PostgreSQL 17
• PeerDB
• Airflow
• Python
• dbt
• Airbyte
• MinIO
• Docker Compose
• GitHub Actions
• Rocky Linux
• nginx
• Nagios
• ELK Stack
• TLS
• HashiCorp Vault
• Data Vault 2.0
• Kubernetes
• pgHero
• pgBadger
• pgwatch
• AWS S3

Oct 2019 – Dec 2021 · Nairobi, Kenya

• Gathered EDMS requirements and developed an electronic document management system to track and digitize over 5 million records
• Built the ETL tool used to load digitized documents into the backend system, orchestrated with Apache Airflow
• Installed and optimized PostgreSQL for geospatial data, scaling the database from 30 TPS to 10,000 TPS; implemented PgBouncer load balancing and automated backup/restore scripts
• Developed an automatic PostgreSQL installation and configuration tool for rapid recovery in case of failure
• Built a desktop application for end users and management to interact with the ETL tool

• Apache Airflow
• PostgreSQL
• ClickHouse
• Nagios
• SQL
• Python

Oct 2025 – Present · Independent Project

• Shortlisted from 2,467 proposals across 5,656 national registrants — NIRU National AI Hackathon 2025, Governance & Policy track
• Architected a 3-node Ollama GPU inference cluster (RTX 3050s) running Qwen2.5:7b — fully self-hosted with no external API dependency
• Built a LangGraph agentic RAG pipeline: Planner → Retriever → Responder with IRAC legal prompting and streaming NDJSON responses
• Implemented semantic caching (Qdrant + Redis, cosine > 0.95, 30-day TTL) to reduce redundant inference
• Qdrant vector search across Kenya Law Reports with a Neo4j citation graph (CITES, OVERRULES, DISTINGUISHES relationships)
• FastAPI microservices with JWT RBAC, Prometheus metrics, and OpenTelemetry tracing
• M-Pesa STK Push payments via Safaricom Daraja API with a FastAPI callback on Railway
• Next.js streaming frontend with admin panel, ingestion job monitor, and health dashboard

• Python
• FastAPI
• LangGraph
• Ollama
• Qdrant
• Neo4j
• PostgreSQL
• MinIO
• Next.js
• Docker Compose
• Redis

Mar 2022 – Nov 2022 · Kenya (concurrent with ArdhiSasa engagement)

• Led full migration of all data pipelines from Xplenty to Prefect orchestration on AWS EC2 — achieving centralized logging and significant cost reduction
• Wrote dbt transformations following the Kimball dimensional modelling approach for an enterprise Snowflake data warehouse
• Built a CI pipeline from scratch in GitHub Actions for automated dbt model testing and fast feedback for contributors
• Introduced SQLFluff for automated SQL quality enforcement, improving team delivery speed by 80%

• dbt
• Snowflake
• Prefect
• Python
• SQL
• GitHub Actions
• AWS EC2
• SQLFluff

Jan 2016 – Dec 2018 · Nairobi, Kenya

• Built a chatbot system for back-office customer support using TensorFlow, deployed on a microservices architecture
• Developed and launched an online academic marketplace, growing it to 200+ active users from zero
• Designed deployment architecture using microservices principles for scalability and maintainability

• Python
• TensorFlow
• SQL
• Microservices

Nov 2015 – Sep 2016 · Nairobi, Kenya

• Built a smart water meter system on an IoT/NB-IoT platform with bi-directional communication between field sensors and back-office
• Reduced water losses for two water utility companies by 30% through real-time sensor-based incident detection

• Python
• Django REST Framework
• PyQt
• NB-IoT