Miller Juma / Senior Software Engineer

Complex systems.
Meaningful impact.

I build the software, data platforms, and AI systems that turn hard problems into things that work. Currently engineering at Microsoft.

Explore my work Download resume

Software & systemsData engineeringApplied AI

THE PERSON BEHIND THE CODE

William Miller JumaEngineer. Builder. Learner.

From the first prototype
to the production problem.

BUILD / MEASURE / IMPROVE

A little less theory. A lot more impact.

A few numbers behind the work

REAL SYSTEMS. MEASURABLE OUTCOMES.

98%

less manual regression triageMicrosoft

99.99%

data pipeline uptimeMicrosoft / Azure Data Factory

70–80%

lower infrastructure costsSolutech / GKE migration

85%

fewer fraudulent accountsSanifu AI / ML-powered onboarding

01 / Selected work

Built to solve.
Not just to ship.

A selection of production systems and independent experiments. Different disciplines. The same care for what happens after deployment.

MicrosoftData engineering / Systems

Making performance data work harder.

A centralized benchmarking platform bringing millions of records from multiple labs into a usable foundation for engineering analysis.

Azure Data FactoryBlob StorageETL

Inside the work

The challenge: Make benchmarking data across labs easier for engineering teams to access and analyze.

My contribution: Architected the centralized platform and engineered pipelines to ingest, clean, transform, and load benchmarking data into databases.

The outcome: 99.99% pipeline uptime and 20% less query and data-access time for 5+ teams.

Professional work. Implementation and source code are not public.

Independent projectAI / Research engineering

Better answers start with better retrieval.

Co-training a RAG reranker and Llama generator in alternating rounds, using generated answers to teach the retriever what matters.

PyTorchQLoRAGRPOPopQA

Explore CoRAG

Griffin Global TechnologiesSoftware engineering

From scattered planning to one platform.

An internal .NET platform connecting employee allocation, project assignments, and utilization. Built from requirements through production rollout.

.NETPlatform designEnd-to-end delivery

Inside the work

The challenge: Replace disconnected planning tools and a licensed stack with an internal system.

My contribution: Owned requirements with operations and finance, technical design, implementation, and rollout.

The outcome: About $100K in annual savings after the SAGE Intacct cutover.

Professional work. Implementation and source code are not public.

Independent projectApplied AI

Customer support that speaks your language.

A multilingual support agent for East African businesses. Grounded in company FAQs, with a fine-tuned Mistral model running on local hardware.

Mistral-7BLoRAFAISSLangChain

Explore the chatbot

Independent projectData / AI

RAG, without the black box.

A framework-free retrieval pipeline from document loading to grounded answers with citations. Then rebuilt with LangChain to compare the approaches.

Azure OpenAIFAISSLangChainMLflow

Explore RAG from scratch

Independent projectSoftware / AI

Turning an idea into a considered decision.

A proposer, critic, and summarizer turn vague feature requests into structured decision documents through bounded, multi-round deliberation.

PythonAzure OpenAIMulti-agent systems

Explore the agent engine

Solutech LimitedCloud / Systems engineering

Less infrastructure. More headroom.

Migrated core sales automation from per-client DigitalOcean VMs to a multi-tenant GKE platform, using consolidation and autoscaling to cut infrastructure costs by 70–80%.

KubernetesGKEMulti-tenancyAutoscaling

More experiments on GitHub

02 / The journey

Depth comes
from doing.

From lending platforms and national-scale portals to Windows benchmarking and Azure data infrastructure.

The full story, in PDF

Microsoft

CURRENT

Software EngineerSep 2024 – Present

Redmond, United States

Building the platforms behind performance engineering: automated regression triage, reliable benchmarking pipelines, and tools that make performance data useful across teams.

98% less manual regression triage.
Millions of benchmark records across multiple labs.
MCP data access supporting 20+ engineering teams.

AzureData platformsPerformance engineering

Griffin Global Technologies

Senior Software EngineerJan – Sep 2024

Owned an internal .NET resource-planning platform from requirements to production, replacing a licensed stack and delivering about $100K in annual savings.

Solutech Limited

Senior Software EngineerMar 2023 – Jan 2024

Modernized a production SaaS platform: multi-tenant GKE infrastructure, a zero-downtime Laravel upgrade, shared packages, and support tooling for 200+ clients.

ViewTech / SasaPay

Software EngineerOct 2022 – Mar 2023

Built an AML engine, developer portal, and CRM integrations for live payments. Automated onboarding, KYC, and risk scoring to cut manual review time by over 70%.

Kinetic Technology

Software EngineerJan – Jun 2022

Delivered exam platforms for KISEB and EBK, connected operations to Dynamics 365 Business Central, and reduced manual administration.

Sanifu AI

Software EngineerAug 2018 – Dec 2021

Built lending workflows, operational data pipelines, and ML-powered fraud detection. Scaled the Ajira Digital portal beyond 100,000 concurrent users and helped reduce fraudulent accounts by approximately 85%.

03 / Behind the work

Curious by nature.
Practical by design.

I like understanding the whole system: the interface someone uses, the data moving underneath, and the infrastructure keeping it alive.

The interesting problems rarely stay inside one discipline.

Over 8+ years, I've worked across full-stack software, distributed systems, data infrastructure, and machine learning. I've led teams of 10+ engineers, worked with multinational teams, and taken systems from the first requirements conversation to the production incident.

Today, my work at Microsoft focuses on performance engineering and benchmarking infrastructure. Outside that work, I explore retrieval, model fine-tuning, and agent systems by building them.

Always learning

MSc, Computer Science In progress

University of East London

BSc, Computer Technology

Jomo Kenyatta University

Read my cryptography thesis

A toolkit, not a box.

MY WORKING STACK

01 Software & systems

End-to-end products, APIs, distributed systems, performance engineering, and cloud infrastructure.

C# / .NETPythonTypeScriptReactNode.jsLaravelC / C++GoAzureKubernetesDockerTerraform

02 Data engineering

Ingestion, cleaning, transformation, orchestration, and analytical data platforms.

SQLAzure Data FactoryBlob StorageADLSDatabricksSpark / PySparkSynapseMicrosoft FabricAirflowKafkaEvent HubsDelta LakePostgreSQLAzure SQLMySQLSnowflakeBigQueryRedshiftPower BIPandas

03 Machine learning & AI

Model development, fine-tuning, retrieval-augmented generation, evaluation, and AI application engineering.

PyTorchTensorFlowScikit-learnHugging FaceLoRA / QLoRALangChainLangGraphFAISSAzure OpenAIMLflowRAGMCP

04 / What's next?

Good work starts
with a conversation.

Have a difficult engineering problem, a thoughtful team, or an interesting opportunity in software, data, or ML? Let's connect.

jumamiller@yahoo.com

Download my resume GitHub LinkedIn

Complex systems.Meaningful impact.