Independent engineering practice / Hamza Ahmed
Production data platforms, backend infrastructure, applied AI, and automation - designed and shipped end to end by a senior engineer who can move from architecture to implementation without losing the thread.
Selective availability for focused freelance engagements.
Engineering work delivered across environments for
A broad engineering range, one accountable owner
I work across the boundaries where projects usually slow down: data architecture, application code, infrastructure, third-party systems, ML and LLM workflows, observability, and delivery. The goal is not to touch every tool. It is to keep the whole system coherent.
Explore capabilities ↓Selected work
Private product · Analytics infrastructure
Private systemA private B2B analytics platform that turns business-facing definitions into the ingestion, transformation, caching, and execution required to answer them.
Private system · Engineering operations
Private systemA control plane for moving software work from requirement to independently verified handoff across AI coding providers without losing ownership, evidence, or state.
Public reference · Data platform
GitHub ↗A queue-first snapshot and change-data-capture service that preserves event order, durable resume state, complete SCD2 history, and a current PostgreSQL view.
Public reference · Data platform
GitHub ↗A configurable ingestion service for snapshots, live change streams, scheduled collections, and targeted repairs into partitioned object storage.
Public reference · Serverless backend
GitHub ↗An event-driven AWS workflow that detects when sponsored creators go live, starts session monitoring, and records audience, chat, transcript, and mention telemetry.
Public reference · Applied AI
GitHub ↗A registry-driven workflow that screens source material, generates structured content, routes invalid output, and prepares valid results for human review.
Public reference · Generative AI
GitHub ↗A GPU-oriented workflow for adapting SDXL to an existing visual language, generating traceable variations, filtering failures, and handing candidates to designers.
Public reference · ML and product intelligence
GitHub ↗A Databricks and PySpark workflow that converts high-volume community chat into sentiment trends and privacy-conscious aggregate research personas.
The public repositories are anonymized reference implementations based on systems I designed and built. They use synthetic data and contain no client credentials, proprietary configuration, private datasets, or confidential business logic.
Selected outcomes
Results across startup, consulting, product, and enterprise environments.
lower pipeline infrastructure cost after reworking streaming compute, compression, and storage.
less manual response time through agentic routing, dedicated conversations, summaries, and personalization.
fewer abuse incidents after deploying a real-time moderation service into production workflows.
faster Athena queries while reducing cost by 30% through partitioning, bucketing, and compression.
Bring the problem, not a preselected stack
Warehouses, lakehouses, CDC, streaming, ETL/ELT, dbt, Airflow, Spark, data quality, history models, recovery, and migration.
Python, Go, Rust, and TypeScript services; APIs; serverless workflows; internal tools; SaaS connections; deployment and operational automation.
LLM pipelines, RAG, model serving, text-to-SQL, support automation, human-review systems, ML workflows, and multi-agent engineering operations.
Architecture reviews, performance tuning, failure recovery, idempotency, testing, observability, cloud cost reduction, CI/CD, and production handoff.
How the work moves
Small debugging work can stay small. Larger builds get enough structure to make architecture, assumptions, tests, and handoff explicit.
Define the business outcome, current failure mode, constraints, source of truth, and what success must prove.
Choose boundaries, data contracts, failure behavior, and a delivery path before adding unnecessary infrastructure.
Implementation, tests, operational checks, and measurable acceptance criteria move together rather than being separate phases.
Document what exists, what was proven, how to operate it, and where the remaining risks or next decisions actually are.
About Hamza Ahmed
I am a Senior Software, Data & AI Engineer with eight years across startups, consulting, product teams, and enterprise client environments. I have owned systems from ingestion and orchestration through APIs, infrastructure, analytics, ML, deployment, and the workflows people use on top of the data.
Before software and data engineering, I studied physics and mathematics and worked in astrophysics research. That background still shapes the work: define the model, test assumptions, trace evidence, and make reliability part of the design.
Knownbyfew is my independent engineering practice. Martian Bee is the product company where I am combining the same disciplines into a private analytics platform.
Start with the problem
A failing pipeline, an unfinished platform, a hard integration, an AI workflow that needs production discipline, or a system that has outgrown its first architecture.