Skip to content

Integrations & LLM agent tooling

FaresKhanchouch

Forward Deployed Engineer

I take changes to banks’ credit platforms from scoping to go-live, connect those platforms to the systems banks already run, and build LLM tooling that sticks to the facts.

See my workRésumé
banks I work with in Qatar & Saudi Arabia
3
production releases
2
freelance clients, solo
6+
records in a data platform I built
1.1M+
Portrait of Fares Khanchouch

Now

Integration Consultant

Axe Finance

Experience

Delivery and integration for banks

Integration Consultant · Axe Finance

Dec 2025 – Present · Tunis, Tunisia

  • Delivery, end to end

    Credit-platform changes for banks in Qatar and Saudi Arabia, from scoping through UAT to go-live and support, including 2 production releases.

  • Core-banking integration

    REST services in C# that feed customers, limits and outstandings into credit workflows, plus SMS alerts secured with OAuth, API keys and IP whitelisting.

Alongside delivery: Internal Python tooling (MCP servers and Claude skills) that explains the credit platform’s configuration and cites its sources.

Career line

Timeline: engineering degree at ISTY from 2021 to 2024; internships at WAY2CLOUD in 2022 and 2023 and at Nuage Up in 2024; freelance automation work from September 2024 to November 2025; Integration Consultant at Axe Finance since December 2025.

Before

  1. Sep 2024 – Nov 2025

    Freelance Automation Engineer · Independent

    Automation projects for 6+ clients as a solo engineer, mostly in n8n and Make: KYC document pipelines with LLM/OCR extraction for an ISO-compliance consultancy, a 100,000+ record Excel-to-Airtable migration via API scripts, and an AI content pipeline publishing ~50 posts a week.

    • n8n
    • Make
    • Airtable
    • LLM APIs

  2. May – Aug 2024

    Cloud & DevSecOps Intern · Nuage Up

    AWS in Terraform, CI/CD with security scanning, Kubernetes deployments with network policies and RBAC.

    • AWS
    • Terraform
    • Kubernetes
    • CI/CD

  3. May – Aug 2023

    Full-Stack Web Development Intern · WAY2CLOUD

    An e-commerce platform in React, Node.js and MongoDB, built solo.

    • React
    • Node.js
    • MongoDB

  4. Jun – Jul 2022

    DevOps Intern · WAY2CLOUD

    A Python REST API automating Kubernetes cluster operations, containerized and deployed with Helm.

    • Python
    • Kubernetes
    • Helm

Projects

Built end to end, and measured

Personal project

Grounded LLM generation & a job-market data platform

A Python MCP server that crawls 63,000+ company job boards through 24 source adapters for applicant-tracking and HR systems (Workday, SuccessFactors, Greenhouse and others) into 1.1M+ records, plus an LLM document generator whose every claim must come from a versioned fact store.

Architecture: job-market data platform and grounded LLM generationIngest: a crawler reads 63,000+ company job boards through 24 applicant-tracking-system adapters, with rate limiting and resumable runs, into a normalized SQLite store of 1.1M+ postings. Generate: a brief (the posting's key terms and requirements plus a menu of facts) goes to the LLM, which picks and rewords facts from a versioned fact vault. Guards check every claim against the vault for invented numbers, invented technologies and cross-document consistency; refused text goes back for a rewrite. Accepted output is rendered to a one-page PDF. An evaluation harness runs AI agents through the real pipeline on 12 benchmark postings and scores the output with blind LLM reviewer panels.IngestGreenhouseWorkdaySuccessFactors+ 21 moreCrawler63k+ boards · rate limits · cachingSQLite1.1M+ postingsnormalizedGenerateBriefkey termsrequirementsFact vaultversionedLLMpicks factsrewordsrefused → rewritechecked againstGuardsnumberstechnologiescross-documentPDFfit to 1 pageEval harness12 benchmark postingsAI agents run itblind LLM panelsscorecard → fix
Posting hard requirements addressed
54%to100%
Bullets judged overclaimed
6%to0%
Blind LLM-reviewer résumé score (1–10 rubric)
5.1to6.3/10
  • Every generated claim traces to a versioned fact store; invented numbers and technologies are rejected.
  • An agent-in-the-loop evaluation harness: AI agents run the real pipeline and blind LLM reviewer panels score the output.
  • Python
  • MCP
  • SQLite
  • Embeddings
  • LLM evals
Read the write-up

Open source

n8n Kubernetes Operator

Runs a complete n8n instance on Kubernetes from one declarative YAML file instead of 8+ hand-written manifests.

Architecture: n8n Kubernetes operatorOne custom resource (replicas, n8n version, PostgreSQL, service or TLS ingress) is applied to the cluster. The Go operator's reconcile loop observes, compares and acts: it creates and updates n8n, PostgreSQL, secrets, persistent storage, a service and an optional TLS ingress in the instance's namespace, recreates anything deleted, and reports status (Ready, N8nReady, PostgresReady, access URLs) back on the custom resource.Custom resourceone YAML filereplicasn8n versionPostgreSQLservice / TLS ingressapplyOperator (Go)observecompareactreconciledeleted resources are recreatedmanagesisolated per instancen8nDeployment · replicasPostgreSQLdatabaseSecretscredentialsPVCpersistent storageServicein-cluster accessIngressoptional TLSstatus: Ready · N8nReady · PostgresReady
To run n8n with PostgreSQL
8+ manifeststo1 resource
  • A custom resource provisions PostgreSQL, n8n, secrets and networking, with isolated multi-instance setups.
  • Self-healing: deleted resources are recreated, and each instance reports its health and access URL.
  • Go
  • Kubernetes
  • Operators
  • n8n
  • PostgreSQL
View on GitHub

About

The short version

I work with client teams from the first scoping conversation to go-live support.

I work directly with banks’ IT, credit and business teams, alongside the on-site delivery team. As a freelancer I did the same for smaller clients, including a paid n8n training that took one client from zero to building their own workflows.

Outside work I build LLM tooling and a Kubernetes operator for n8n.

Based in
Tunis, Tunisia
Education
Engineering Degree in Computer Science · ISTY, Université Paris-Saclay · 2021 – 2024
Languages
Arabic (native)French (C1)English (C1, TOEIC 985/990)

Tools I work with

used in the projects above

AI & agents

  • MCP servers (used in a project above)
  • LLM APIs
  • LLM evals (used in a project above)
  • LLM output guardrails

Integration

  • REST APIs
  • Webhooks
  • OAuth / API keys
  • n8n (used in a project above)
  • Make (automation)

Programming

  • Python (used in a project above)
  • TypeScript
  • C#
  • Go (used in a project above)
  • T-SQL / SQL Server

Platform

  • Docker
  • Kubernetes (used in a project above)
  • Terraform
  • AWS
  • PostgreSQL (used in a project above)

Contact

Say hello

Email is the quickest way to reach me. The form works too.

Email

Tunis, TunisiaUTC+1