A meeting enters the local workflow.
Browser recording or imported audio is stored locally first, then uploaded in recoverable chunks when connectivity permits.
Applied AI & Backend Engineer
I build intelligent systems that turn complex data into useful business actions.
Software Engineering graduate building applied AI, multilingual speech-processing systems, LLM-powered business software, workflow automation, and production-oriented backend platforms.
LIVE SYSTEM MAP5 services connected
Speech ingest AI processing Business action
Engineering AI for real operating conditions
I work across model evaluation, AI service integration, backend architecture, and product delivery. My focus is not AI in isolation, but systems that remain private, observable, recoverable, and useful.
I graduated in 2026 with a Bachelor's Degree in Software Engineering and Information Systems from the Faculty of Sciences of Monastir. My work spans local inference, multilingual speech, queues, APIs, automation, data processing, and full-stack product engineering.
Production-oriented architecture Multilingual systems Local and on-premise AI
SalesAI - Meeting Insights Hub
The centerpiece: a private, on-premise conversational intelligence platform built for multilingual Tunisian sales meetings.
Conversational intelligence / Applied AI / Full-stack platform
From code-switched conversation to a traceable next action.
SalesAI transforms recorded meetings into transcripts, structured commercial insights, client records, follow-up drafts, and sales analytics across Tunisian Derja, French, Arabic, and English.
18.2htraining audio5,451training samples309test samples5,200retained checkpoint
PROCESSING TOPOLOGYLocal-first / queue-backed
01AudioBrowser / import
02WhisperDerja + code-switching
03TranslationSegment-aware NLLB
04LLM analysisGrounded extraction
05ActionsCRM / email / analytics
IndexedDB Sync after reconnectionChunk upload Redis / BullMQ fallbackAudit log Processing diagnostics
Scroll-linked system walkthrough
One meeting.
Four operational states.
Capture
Interpret
Reason
Act
01 / 04CaptureIndexedDB / chunk upload
Speech becomes language-aware evidence.
Whisper processes Tunisian Derja and code-switching while segment-level translation preserves the meeting timeline.
The pipeline extracts commercial structure.
Grounded analysis identifies objections, risks, sales stage, next actions, sentiment, and follow-up context.
Results return as operational state.
Client records, versioned emails, dashboards, diagnostics, and audit events make the output usable and traceable.
System capabilities
Browser recording and chunked audio import
Local multilingual transcription and segmented translation
Summaries, objections, risks, stage, sentiment, and next actions
Client 360, email history, deterministic offer generation
Offline operation with synchronization after reconnection
Queue diagnostics, audit logging, and processing fallbacks
Key engineering decisions
Privacy shapes architecture
Local inference and on-premise deployment keep sensitive meeting audio within the client environment.
Deterministic where it matters
Offer emails use imported source documents and controlled logic rather than ungrounded generation.
Failure is a first-class state
Local persistence, resumable uploads, queue fallback, diagnostics, and audit trails support recovery.
Frontend
React / TypeScript / Vite / Tailwind CSS / shadcn/ui / TanStack Query / Dexie / IndexedDB
Backend
Node.js / Express.js / MongoDB / Mongoose / JWT / RBAC / BullMQ / Redis / Multer / bcrypt
AI services
FastAPI / Whisper Large v3 / faster-whisper / NLLB-200 / Gemini / PyTorch / Transformers / FFmpeg
Engineering
Vitest / Pytest / Docker / CUDA / Queue diagnostics / Audit logging / Fallback processing
Read the full case study
Includes architecture, adaptation methodology, reliability design, limitations, and lessons learned.
Selected projects
A focused set of backend, full-stack, and machine-learning work. Public links are shown only where available.
Marketplace platform
BidEk
An online auction marketplace supporting accounts, listings, live bidding workflows, notifications, and responsive buyer and seller interfaces.
- Contribution
- Full-stack application design and implementation
- Status
- Completed project
Private / link unavailable
Machine learning pipeline
UNSW-NB15 Network Intrusion Detection
An end-to-end multiclass intrusion-detection pipeline with preprocessing, feature engineering, model comparison, confusion matrices, and false-positive and false-negative analysis.
- Contribution
- Data preparation, modeling, evaluation, and analysis
- Status
- Public repository
Backend API
Open Threat Intelligence Hub
A FastAPI backend that aggregates and normalizes threat indicators, calculates risk scores, and exposes filtering, lookup, and enrichment APIs.
- Contribution
- API architecture, normalization, and scoring logic
- Status
- Public repository
Experience
01
Netcom SA
AI Developer / Software Engineer
Designed and developed SalesAI, an on-premise conversational intelligence platform for multilingual sales meetings.
- Built the frontend, backend, local-first synchronization, AI services, processing pipeline, diagnostics, email workflows, and business dashboards.
- Integrated Whisper, faster-whisper, NLLB-200, Gemini, MongoDB, Redis, BullMQ, Express, FastAPI, and React.
- Fine-tuned and evaluated Whisper Large v3 for Tunisian Arabic and code-switched speech, with production constraints around privacy, reliability, and traceability.
02
Netcom SA
AI & Workflow Automation Intern
Designed and deployed maintainable n8n automation workflows in Docker for cross-team operations.
- Automated approvals, notifications, and synchronization between design, production, and finance teams.
- Added validation, structured logging, monitoring, access-control checks, error handling, and recovery documentation.
Technical range
Grouped by engineering domain, without subjective ratings.
Programming
PythonJavaJavaScriptTypeScriptBashSQLBasic C
AI & Data
Whisperfaster-whisperPyTorchTransformersscikit-learnPandasNumPyNLLB-200LLM integrationWER/CER evaluationSpeech pipelines
Frontend
ReactNext.jsTypeScriptViteTailwind CSSBootstrapThymeleafshadcn/ui
Backend
Node.jsExpress.jsSpring BootFastAPIREST APIsJWTRBACMongoDBPostgreSQLSQLite
Infrastructure & Tools
GitGitHubDockerRedisBullMQn8nLinuxWSLMavenIndexedDBDexieCUDAFFmpeg
How I approach engineering
01
Frame the operation
Start with the decision, constraint, or workflow the software must improve.
02
Separate evidence from inference
Keep data preparation, model experimentation, and production services independently testable.
03
Design the failure path
Build for observability, recovery, fallbacks, and useful behavior when dependencies fail.
04
Make outputs traceable
Prefer grounded, inspectable results and deterministic logic where generation adds risk.
Education
Faculty of Sciences of Monastir
Bachelor's Degree in Software Engineering and Information Systems
2023 - 2026 / Graduated 2026
Artificial Intelligence / Machine Learning / Data Science / Backend Engineering / Software Architecture / Information SystemsCertifications
Aviatrix Certified Engineer - Multicloud Network Associate
AWS EC2/S3 Certificate
Linux Certificate
Cisco Networking Academy - MEA&CIS Cybersecurity Essentials
Certified Cybersecurity Educator Professional
Languages
ArabicNative
EnglishC2
FrenchC1
GermanB1
Available for new work
Let's build an AI system that survives contact with reality.
I am open to applied AI, backend engineering, workflow automation, machine learning, and full-stack opportunities.
I am especially interested in teams building real AI-powered products, multilingual systems, automation platforms, or scalable backend services.
nederhassen2@gmail.comRemote international / hybrid or onsite in Tunisia / full-time / part-time / contract / freelance / alternance