Information Technology · Full-time
Direct employerSenior AI Engineer

Published by employer
About this role
This plain-English summary was prepared by workegypt.africa from the original advert. Confirm all details on the source before applying.
Role overview
Thndr is hiring a full-time Senior AI Engineer for its Cairo office in Egypt. The role focuses mainly on Alpha, an AI agent that prepares customer orders for confirmation, and suits an experienced engineer who can take AI systems from research and design through to production.
Read full role
What you will do
- Own Alpha modules from specification to production, including their behaviour, production quality and long-term maintenance.
- Design tool contracts, model context, capability routing and code-based guardrails around customer needs.
- Build trusted observability and evaluations by analysing traces, classifying failures, checking resulting states and validating LLM judges against human labels.
- Manage latency, reliability and cost for your area, including streaming, provider limits, fallbacks and production alerts.
- Engineer retrieval, planning, allocation, event-processing and money-service integrations with idempotency, retries and reconciliation.
- Evaluate, fine-tune, serve and safely release transformer models for Alpha and other Thndr products.
- Lead cross-squad designs, investigate questions using production data and experiments, review colleagues’ work, improve coding-agent practices and prototype AI solutions for other products.
Requirements
- Have strong production software engineering experience, including building and operating backend systems and understanding slow dependencies, retries, duplicate events and deployments during active processing.
- Experience with Python, FastAPI, PostgreSQL and queues is preferred.
- Have delivered an LLM agent or tool-calling system to real users and be able to explain decisions about context, tool descriptions, routing, failures and possible improvements.
- Have operated an LLM-backed service, covering time-to-first-token and other latency budgets, streaming, timeouts, fallbacks, costs and incidents.
- Have created evaluations from real traces and understand the limits of single successful runs and misleading metrics.
- Have practical experience fine-tuning and evaluating transformer models using SFT or LoRA, then moving them from training to monitored serving with version control.
- Be comfortable using SQL and Python for analysis, designing experiments and defending conclusions.
- Take responsibility for outcomes across the full stack, including incidents, monitoring and follow-up fixes, and challenge an incorrect brief when necessary.
- Use coding agents daily, verify their output and be able to explain how you organise the work.
- Write clear design documents and findings that people outside engineering can understand and act on.
- Have experience across more than one area, such as agents, retrieval, models, data or backend, and learn unfamiliar areas quickly.
- Be able to decide what is worth exploring and when an experiment should end.
How to apply
Apply on the employer's application page. The employer has not published a closing date.