AI systems
Evaluation-first
LLM-as-judge, NLI, RAG, guardrails
Available for high-impact work Tunis · Remote worldwide
Software engineer · AI engineer
Ismail
Khlifi.
I engineer production software and AI workflows that turn complex systems into clear, dependable products.
LLM evaluation, human-in-the-loop automation, and full-stack platforms—from model logic to the interface people trust.
Current focus
Reliable agentic systems with measurable evaluation.
Engineering degree
ESSAI · 2026
01Approach
Built at the intersection of intelligence and usability.
I connect model behavior, backend architecture, structured data, and product experience into one coherent system.
My work goes beyond wiring an API to a polished interface. I design the feedback loops, failure paths, and review states that make AI features useful in production.
That same rigor shapes my full-stack work: clear data models, maintainable services, accessible interfaces, and performance decisions that hold up beyond the demo.
Open to opportunitiesFreelance available
Workflows
Human-centered
Review loops, routing, approvals
Engineering
End-to-end
React, FastAPI, PostgreSQL, Docker
Foundation
Statistical
Engineering degree · ESSAI 2026
02Experience
From model research to production systems.
Hands-on roles across AI evaluation, workflow automation, full-stack engineering, and applied data science.
01MandaNetwork
Data Science Intern / AI Full-Stack Developer
Feb 2026 - Aug 2026
Tunisia
- Built AI-enabled workflow and evaluation features with React, TypeScript, FastAPI, PostgreSQL, SQLModel, Alembic, and Docker
- Developed human-in-the-loop review flows with AI checks, fallback review, accept/reject decisions, and reviewer comments
- Designed annotation and content-generation workflows that supply LLMs with context, evidence, model answers, and evaluation criteria
- Automated task routing, dependencies, approvals, notifications, capacity tracking, and early issue flagging
- Researched prompt engineering, LLM-as-judge, NLI framing, benchmarks, and failure cases; raised an accessibility audit score from 5.7 to 9.4
FastAPIReactPostgreSQLLLM EvaluationDocker
02Wisecool
Full Stack Engineer
Jul 2025 - Aug 2025
Tunisia
- Built an installment subscription payment system with Next.js and Django, cutting API response time by 30%
- Improved calendar, live events, attachments, and a public library for a 500+ user production platform
Next.jsDjangoPostgreSQLREST APIsFull Stack
03Keepcom
Data Scientist / Web Developer
Jun 2024 - Jul 2024
Tunisia
- Analyzed 167+ Facebook & Instagram ad campaigns data to measure ROI and optimize targeting
- Built regression, clustering, and time-series models with 80%+ accuracy
- Built a company reporting site with Node.js and Express.js, reducing manual reporting effort by 30%
PythonScikit-learnExpress.jsNode.jsData AnalysisML
04DrugIT
ML Engineer & Full-Stack Developer
Jul 2024 - Aug 2024
Tunisia
- Set up Docker environments and CI/CD workflows for a 5-person team, cutting setup time by 40%
DockerCI/CDMachine LearningPython
03Education
A quantitative foundation for intelligent systems.
Engineering, statistics, and applied mathematics shape how I reason about software and model behavior.
2023 - 2026Degree awarded
Engineering Degree in Statistics and Information Analysis
ESSAI, University of Carthage
2021 - 2023
Preparatory Cycle in Mathematics
IPEIEM, El Manar
04Capabilities
A focused stack for shipping complete products.
Selected tools grouped by the problems they solve—not an endless logo wall.
01
AI & Agentic AI
LLMsPrompt EngineeringLLM-as-judgeAI EvaluationRAGFallback LogicGuardrailsHuman-in-the-Loop
02
Backend
DjangoFastAPIPythonSQLModelREST APIsAlembicRedisCelery
03
Frontend
ReactNext.jsAngularTypeScriptTailwind CSSFramer MotionRBAC Dashboards
04
Database & Cache
PostgreSQLSupabaseChromaDB
05
DevOps & Tools
DockerGitGitHub ActionsCeleryLinuxCI/CDn8nRailwayVercel
06
Others
NLPNLIPyTorchTensorFlowscikit-learnWorkflow AutomationAccessibility
05Selected work
Products with depth, not just polish.
AI platforms, developer tooling, and production web experiences—designed and engineered end to end.
01
Live
02
Live
03
Case studyIn development
DevLog - AI Development Journal
Full-stack platform that captures GitHub commits, processes them into coding sessions, and generates AI narratives using GPT-4
Next.js 15Django 5PostgreSQLRedisCelery
04
LiveIn development
05
Live
06
Case study
ChurnPredictor - Customer Analytics
AI-powered customer churn prediction platform with real-time analytics dashboard and 85%+ ML accuracy
Next.js 15TypeScriptFastAPIscikit-learnMetabase
06 · Start a conversation
Have a complex idea worth making simple?
I'm open to AI engineering, full-stack product work, and collaborations where thoughtful systems matter.