TransactAPI
Concurrency-safe financial core in Spring Boot.
- Pessimistic row-level locking
- Idempotency keys for safe retries
- PostgreSQL · JWT · Spring Security
Outcome
0 double-spends under contention
Nairobi, KE · 1.2921° S, 36.8219° E
Darlene Wendy
Nasimiyu
Engineering taught me how to read physical systems and reason around it. Now I write software and train models that control systems.
Software & Machine Learning engineer building high-integrity backends, applied ML systems, and industrial IIoT pipelines from embedded firmware to cloud.
Technical core
Languages
Java, Python, C++, TypeScript, SQL
Backend
Spring Boot, FastAPI, Flask, Django, Spring Security, Spring AI, WebFlux
ML / AI
PyTorch, TensorFlow, Scikit-learn, XGBoost, LangGraph
Data / Infra
PostgreSQL, TimescaleDB, Redis, MQTT/HiveMQ, Docker, AWS (EB, EC2)
Embedded
ESP32-S3, FreeRTOS, C++
Section_01
Backend · high-integrity architectureSoftware Engineering
View repo →
JazaCharge
AI-assisted battery-swap coordination for e-boda and e-matatu fleets over USSD/SMS.
- Africa's Talking USSD / SMS gateway
- Spring Boot orchestration + queueing
- Hackathon origin, still iterating
Outcome
Feature-phone reachable
Leak Detection Pipeline
End-to-end IIoT backend for copper tailings pipelines: telemetry ingest, orchestration, and inference fan-out.
- Spring Boot API + Flask ML service
- ESP32-S3 firmware → cloud ingest
- AWS deployment, React/TS operator UI
Outcome
Embedded → cloud, end-to-end
Concrete Compressive Strength Predictor
XGBoost regression served through a FastAPI inference layer and a Spring Boot client, with a live prediction endpoint.
- XGBoost regression pipeline
- FastAPI inference · Spring Boot client
- Live prediction endpoint
Outcome
R² = 0.9407
Section_02
Applied modeling · physics-informedMachine Learning
LSTM Leak Classifier
Physics-informed sequence model for pipeline anomaly detection on 51 engineered features.
Darcy-Weisbach pressure modeling drives the feature space.
99.81%
Accuracy
100%
Leak recall
51
Features
AISI 316 Flow Stress ANN
Neural surrogate for hot-compression flow stress in stainless steel: closed-form replacement for empirical constitutive models.
Trained with scipy L-BFGS-B against experimental hot-compression data.
0.9886
R²
L-BFGS-B
Optimizer
Concrete Compressive Strength Predictor
XGBoost regression on standard concrete mix + curing features, served for live inference.
Feature engineering on age, cement, water, aggregates and admixtures.
0.9407
R²
XGBoost
Model
Section_03
Agentic systems · industrial researchAI Engineering & Research
ResearchMind - In Progress
Multi-Agent Orchestration
A LangGraph-driven research assistant orchestrating specialized Claude agents. FastAPI shell, Redis pub/sub for inter-agent messaging, Tavily for grounded web retrieval.
LangGraph · FastAPI · Redis · Claude · Tavily
Olkaria II Geothermal Optimization
Research Collaboration
Multi-objective plant optimization using NSGA-II with an ANN surrogate model over IAPWS-IF97 steam thermodynamics. Targeting throughput / efficiency Pareto frontiers at Olkaria II.
NSGA-II · ANN Surrogate · IAPWS-IF97View repo →
Section_04
Notes · essays · MediumWriting
Medium · 10 min read
From Raw Data to API: Building a Concrete Strength Prediction Pipeline with XGBoost, FastAPI, and Spring Boot
Most machine learning tutorials stop the moment the model trains successfully — this one goes further, from notebook to a real deployed API.
Read on Medium →Medium · 8 min read
Scaling Systems 101: Vertical Scaling, Horizontal Scaling, Load Balancing, and Database Replication
Part 1 of a beginner-friendly system design series for backend, ML, and AI engineers.
Read on Medium →Section_05
Stack · tools · runtimesTechnical Skills
Languages
- Java
- Python
- C++
- TypeScript
- SQL
Backend
- Spring Boot
- FastAPI
- Flask
- Django
- Spring Security
- Spring AI
- WebFlux
ML / AI
- PyTorch
- TensorFlow
- Scikit-learn
- XGBoost
- LangGraph
Data / Infra
- PostgreSQL
- TimescaleDB
- Redis
- MQTT / HiveMQ
- Docker
- AWS (EB, EC2)
Embedded
- ESP32-S3
- FreeRTOS
- C++
Section_06
Roles · engagementsExperience
- Analytics LeadSoko Sauti · AI marketing SaaS
Built the 12-month financial model used in the pre-seed investor meeting.
- Freelance Backend DeveloperFiverr
Spring Boot and FastAPI project delivery for external clients.
Section_07
Formal · appliedEducation
- 2026BSc Mining and Minerals Processing EngineeringJKUAT
- Software Engineering - Backend SpecializationALX / Holberton
- Data Science and Machine LearningALX, JKUAT
Note_00 · Engineering philosophy
The through-line.
Mining engineering got me through the door of university. It wasn't where I found my footing; coding was. Somewhere in year three I stopped forcing myself through mineral processing problem sets and started staying up late building things instead, and that was the first time engineering felt like mine.
I still carry the rigor from that background: physics-informed features, deterministic systems, respect for what happens when something fails in the real world. But I don't want a mining degree to be the headline of who I am. I'm a software and ML engineer now. That's the through-line I'm building on.
Transmission
Available for collaboration.
Open to backend, ML, and applied-research roles and even Collaborations.