Portrait of Taha Shalaby

One mind,
full spectrum.

I build production AI systems and teach Cambridge Mathematics and Computing. Different wavelengths, same light.

Open to engineering and teaching roles

Taha-Yassen Shalaby, Full-Stack AI Engineer and Mathematics & Computing Educator

6+

years shipping production AI

7+

years teaching Cambridge curriculum

12+

ML systems for Fortune 500 clients

2M+

predictions served per month

Taha Shalaby

Two careers, one feedback loop.

Most engineers can build. Most teachers can explain. I have spent the last decade doing both at once, and each career keeps making the other one better.

On the engineering side, I co-founded an AI startup that shipped 12+ production ML systems for Fortune 500 clients, led validator infrastructure for a blockchain oracle network, and now build agentic LLM applications for cybersecurity: on-prem tool-calling agents that let SOC analysts query security logs in plain English.

My students learn computing from someone who deployed a model to production the night before.

In the classroom, I have spent over seven years as a Cambridge Mathematics and Computing teacher across international schools in Cairo, Egypt, from Year 1 Computing through Checkpoint Mathematics. The two careers sharpen each other: my clients get systems explained in plain language, and my students get lessons grounded in a live industry.

Portrait of Taha Shalaby
Taha Shalaby. Cairo, Egypt. U.S. Citizen.

Systems that shipped.

Six years of production AI, from startup CTO to agentic LLM applications for security operations.

Snappers, AI Cybersecurity

May 2025 to Present

Current

React / FastAPI / PostgreSQL / LangChain / Ollama / Qwen / Docker Compose

AI Engineer / Consultant

Building agentic LLM applications for security operations: an AI-powered SOC analyst dashboard with a tool-calling agent grounded in live log data.

  • Built a tool-calling LLM agent (on-prem Qwen via Ollama) that lets Tier-1 SOC analysts replace hand-written SIEM queries with plain-English requests, cutting query construction from minutes to a single step.
  • Hardened the agent against OWASP LLM Top 10 risks: validated generated queries, read-only least-privilege execution, full prompt and response audit logging.
  • Built an automated report engine for PCI compliance and incident reports with modular, customizable templates.
  • Deployed fully on-prem with open-source models to meet strict data residency requirements.

Singularity Finance

Feb 2023 to Apr 2025

Node.js / TypeScript / FastAPI / GCP MIG / Ethereum

Tech Lead

Led a small dev team building and operating price-feed algorithms for the Flare FTSO oracle network.

  • Landed submissions within the reward band on roughly 90% of price windows on the Songbird canary network.
  • Architected and operated infrastructure on GCP Managed Instance Groups, sustaining about 99% uptime across a 3-node validator fleet.
  • Implemented Node.js/TypeScript backend services integrated with FastAPI Python endpoints for ML algorithms in an Ethereum-based ecosystem.
  • Managed confidential VMs, automated snapshot and backup processes, and ERC token node infrastructure.

laccuna, AI / Data Science Startup

Dec 2019 to Dec 2022

PyTorch / Transformers / FastAPI / GCP / Ray / Kubernetes

Co-Founder & CTO

Co-founded and set technical direction for an AI startup shipping production ML for SMEs and Fortune 500 clients.

  • Delivered 12+ production ML projects powering 2M+ predictions per month across manufacturing, media, finance, and telecommunications.
  • Led a consulting engagement with UCI graduate engineers for a multi-million-dollar California manufacturing company.
  • Built production systems spanning NLP, time-series forecasting, classification, deep learning, and distributed computing.
  • Owned backend architecture, API development, GCP deployment, and the client delivery pipeline end to end.

Selected work

Snappers, 2025

AI-Powered SOC Dashboard with Agent Chat

Problem

Tier-1 SOC analysts were hand-writing SIEM queries to investigate incidents, and strict data residency requirements ruled out cloud LLMs entirely.

Approach

A full-stack dashboard with a tool-calling agent running Qwen on-prem via Ollama, grounding answers in live log data, validating every generated query against read-only, least-privilege access, and automating PCI and incident reporting.

React / FastAPI / PostgreSQL / Ollama / Qwen / Docker

Outcome

Query construction dropped from minutes of manual syntax to a single plain-English request, fully on-prem.

laccuna, 2021

BERT Article Relevance Classifier

Problem

A US media company screened article relevance with around 100 manual analysts.

Approach

Fine-tuned a BERT transformer classifier with PyTorch and HuggingFace.

82%

F1-score, with analyst headcount cut from 100 to 15.

PyTorch / HuggingFace / BERT

laccuna, 2022

Distributed Time-Series Forecasting

Problem

Client-specific label forecasting had outgrown single-machine training.

Approach

A Ray-parallelized pipeline across a GCP compute cluster with interactive Plotly dashboards.

OutcomePortfolio-wide forecasts delivered at cluster scale for a US media company.

Ray / PyTorch / statsmodels / GCP / Plotly

laccuna, 2022

Customer Churn Prediction

Problem

A telecom client was losing customers with no early warning signal.

Approach

A tuned gradient boosting model deployed on GCP and wired into retention workflows.

15%

reduction in customer churn.

XGBoost / scikit-learn / GCP

laccuna, 2022

ML Workflow Engine on Kubernetes

Problem

Every client deployment meant rebuilding orchestration from scratch.

Approach

An MLOps microservices workflow engine on GKE with auto-scaling, ingress, secrets management, and configurable Dockerfiles.

2M+

predictions per month across the full client portfolio.

FastAPI / Celery / RabbitMQ / Nginx / GKE

A hand writing calculus formulas in chalk on a blackboard
Cambridge Mathematics and Computing, Years 1 to 8.

Photo 1 of 4: A hand writing calculus formulas in chalk on a blackboard

Credentials

  • IPC Level 1: Implementing the International Primary Curriculum
  • Cambridge International: 7+ years delivering Checkpoint & IGCSE Mathematics and Computing

Seven years at the whiteboard.

Cambridge International Mathematics and Computing across Cairo's international schools, currently at Wycombe Abbey Cairo East.

Mathematics & Computing Teacher

Sept 2025 to Present

Wycombe Abbey Cairo East

Delivering Cambridge Mathematics to Years 7 and 8 Checkpoint students and Computing across Years 1 to 8.

  • Prepare Checkpoint students for comprehensive end-of-stage assessments with differentiated, inquiry-based lesson design.
  • Teach digital literacy, computational thinking, and programming across Primary and Lower Secondary.
  • Bring active industry AI experience into Computing lessons for authentic, contemporary relevance.

Homeroom Teacher, Milepost 3

Jul 2024 to Aug 2025

Kompass Education

Delivered English, Mathematics, Science, and Health through the IPC framework to students aged 9 to 12.

Mathematics Educator

Sept 2023 to Jun 2024

New Capital English School

Taught Cambridge Mathematics for Years 6 to 8 and completed Checkpoint examinations for Years 6 and 8.

Mathematics Educator

Sept 2018 to Sept 2020

New Capital English School

Developed Cambridge-aligned Mathematics curriculum for lower secondary Checkpoint and pre-IGCSE.

ICT Educator

Sept 2017 to Sept 2018

Futures British School

Delivered the ICT curriculum for lower secondary and pre-IGCSE, aligned with Cambridge standards.

Two tracks, one decade.

Teaching since 2017, shipping AI since 2019, and for most of it, both at once.

2014 to 2017

Learning

A mechatronics degree at AAST, ending with an autonomous attendance robot that used facial recognition.

  • B.Sc. Mechatronics Engineering, AAST, Cairo2014 to 2019

2017 to 2020

Both at once

First classroom and a robotics club, then Cambridge maths. Co-founded laccuna in late 2019 while still teaching.

  • B.Sc. Mechatronics Engineering, AAST, Cairo2014 to 2019
  • ICT Educator, Futures British SchoolSept 2017 to Sept 2018
  • Mathematics Educator, New Capital English SchoolSept 2018 to Sept 2020
  • Co-Founder & CTO, laccuna, TorontoDec 2019 to Dec 2022

2020 to 2023

All in on AI

Three years as laccuna's CTO shipping 12+ production ML systems, then tech lead on Flare oracle infrastructure.

  • Co-Founder & CTO, laccuna, TorontoDec 2019 to Dec 2022
  • Tech Lead, Singularity FinanceFeb 2023 to Apr 2025

2023 to now

Both again

Back at the whiteboard in 2023. Now agentic AI at Snappers alongside Maths and Computing at Wycombe Abbey.

  • Tech Lead, Singularity FinanceFeb 2023 to Apr 2025
  • Mathematics Educator, New Capital English SchoolSept 2023 to Jun 2024
  • Homeroom Teacher, Kompass EducationJul 2024 to Aug 2025
  • AI Engineer / Consultant, Snappers, AI CybersecurityMay 2025 to Present
  • Mathematics & Computing Teacher, Wycombe Abbey Cairo EastSept 2025 to Present

The toolkit.

As an engineer

AI / ML

  • PyTorch
  • HuggingFace Transformers
  • LLM agents & tool calling
  • RAG
  • LangChain
  • Ollama
  • LLM security (OWASP LLM Top 10)
  • scikit-learn
  • XGBoost
  • NLP

Full-Stack

  • Python
  • TypeScript
  • React
  • Next.js
  • Tailwind CSS
  • FastAPI
  • Node.js
  • PostgreSQL
  • SQL

Cloud / DevOps

  • GCP (GKE, Cloud Run, MIG)
  • AWS
  • Docker
  • Kubernetes
  • MLOps
  • CI/CD
  • Linux

AI Tooling

  • Context engineering
  • MCP (Model Context Protocol)
  • Agent harnesses: skills, hooks, subagents
  • Multi-agent orchestration

As an educator

Curriculum

  • Cambridge Checkpoint
  • Cambridge IGCSE
  • IPC Framework
  • Pre-IGCSE Mathematics & Computing

Pedagogy

  • Inquiry-based learning
  • Differentiated instruction
  • Formative & summative assessment
  • Cross-curricular design

EdTech

  • GeoGebra
  • Desmos
  • Wolfram Alpha
  • Google Classroom
  • Scratch
  • Python in the classroom

Get in touch.

Hiring for an AI engineering role, building out a maths or computing department, or something in between: I would like to hear about it.

tahashalaby93@gmail.com
I'm reaching out about