Lorenc Leci

Design and ship bespoke GenAI products end-to-end — from client discovery through production — for clients including the Mediasoft ERP ecosystem.

• Built WivenLLM, a production-oriented hybrid-RAG assistant for the Mediasoft ERP ecosystem (FastAPI · MongoDB · React), grounding every answer strictly in official documentation to reduce hallucination.
• Engineered a 7-crate Rust ontology graph-database with its own append-only storage engine and a hybrid lexical + vector + graph-expansion RAG pipeline (axum HTTP API).
• Co-designed wiven-os, a shared multi-agent orchestration OS reused as the safety/governance foundation across every client project.
• Led forward-deployed client engagements (My Traiteur, APC) from discovery to prototype, with human-in-the-loop guardrails and adversarial evaluation harnesses.
• Owned DevOps and production-readiness: Docker, GitHub Actions CI/CD, and provider-agnostic LLM deployment (OpenAI, Anthropic, NVIDIA NIM, DeepSeek, Ollama, vLLM).

• Python
• Rust / axum
• FastAPI
• Hybrid RAG
• Multi-agent
• React / Next.js
• MongoDB
• Docker

Automated threat-intelligence workflows and built reliable data pipelines for noisy malware-analysis data.

• Processed malware-analysis datasets and automated repeatable threat-intelligence workflows with Python, Linux scripting, and CAPEv2.
• Designed data-processing pipelines for noisy security data: extraction, normalization, quality checks, and reproducible analysis.

• Python
• ML / data pipelines
• CAPEv2
• Linux

Built and evaluated an NLP sentiment-analysis pipeline for social-media data.

• Built an NLP sentiment classifier with scikit-learn, spaCy, TF-IDF and embeddings, with SQL-backed dataset preparation.
• Strengthened data-engineering fundamentals: cleaning, train/eval splits, and model-quality analysis.

• Python
• NLP
• scikit-learn
• spaCy
• SQL

Applied compiler techniques and metaheuristics to code-obfuscation research.

• Applied LLVM, JMetalPy, API integration and database tooling to code-obfuscation research.
• Built foundations in reading complex codebases and documenting technical trade-offs.

• LLVM
• JMetalPy
• APIs
• Databases