Freelancer profile translated to English.
Back to original languageAbout Yaya
LLM & Speech AI Engineer
I design and deployoptimized LLMs, intelligent chatbots, RAG systems on your data, and speech generation solutionsfor natural, multilingual, and efficient AI interactions.
💬 Conversational AI & LLM
✅AI Chatbot Developmentbased on closed LLMs (Claude, GPT, Gemini, etc.) or open-source LLMs (Mistral, Qwen, DeepSeek, etc.)
✅RAGon your internal documents (PDF, CRM, SQL database, etc.) withLangChain, LlamaIndex, FAISS, Pinecone, ElasticSearch, etc.
✅ **LLM Deployment**: quantization (4-bit, 8-bit) or distillation to reduce costs and latency of LLMs deployed on servers of your choice or locally.
✅Embeddings & Semantic Search
✅Document Extraction and Structuring(PDFs, Scans, etc.) to extract insights and value.
🎙️ Voice AI & Speech Processing
✅ High-precisionSpeech-to-Text(Whisper, MMS, Wav2Vec2, etc.) for video/audio transcription and translation.
✅Speech Synthesis and Voice Cloning(ElevenLabs, open-source models) for voice assistants, podcasts, video translations
🔧 Stack & Tools
Python, PyTorch, Hugging Face, LangChain, LlamaIndex, FAISS, Pinecone, GPT API, Athropic API, Docker, AWS, GCP, Azure
Feel free to contact me, I would love to discuss your needs and explore how I can contribute to the success of your project.
French
Native or bilingual
English
Fluent
Fula
Native or bilingual
Remote only
Primarily works remotely
Experience
- AvocalixDevelopment of conversational agents using a proprietary LLMSOFTWARE PUBLISHINGOctober 2023 - December 2023 (2 months)Paris, FranceI worked as a Freelancer to develop conversational agents specialized in law, using the RAG method. These LLM-based conversational agents and Chatbots must be able to meet user needs by using a large database of legal texts, and many other tools to solve the tasks requested by users. For reasons of user data sensitivity, the chatbots had to use an LLM deployed by ourselves, on servers that we control, and not use OpenAI's APIs. The whole challenge is to deploy a lightweight LLM, inexpensive in computation and fast to respond to many requests. As for the chatbots using this LLM, they must be as accurate as possible by using a database of legal texts and know how to use the right tools at the right time. I developed these LLM agents based on Langchain. In some cases Langchain was not enough, so I switched to Guidance.
- CNRSNLP EngineerRESEARCHDecember 2022 - April 2023 (5 months)Paris, FranceProcessing of voice data using NLP and Speech Processing methods. The main work consisted of training Language Models and Speech Models on large amounts of voice data. The projects were entirely coded in Python using the PyTorch library for model development, and the Transformers library (Hugging Face) for pre-trained NLP models. The models used are: HuBERT (BERT for speech), CPC, Whisper, and classic language models.
- Ecole Normale SupérieureNLP EngineerRESEARCHSeptember 2022 - December 2022 (4 months)Paris, FranceDevelopment, training, and evaluation of self-supervised models on voice data (self-supervised speech models) and Language Models. The work was done entirely in Python using Pytorch and the Fairseq NLP library.
Reviews
Recommendations
These freelancer profiles also match your criteria
Agatha Frydrych
Backend Java Software Engineer
4.7
(3)
2
Baptiste Duhen
Fullstack developer
4.6
(4)
5
Amed Hamou
Senior Lead Developer
4
(2)
7
Audrey Champion
Web developer
4.3
(3)
4
Education
- MasterUniversité Paris-Cité2022Master en NLP: Traitement Automatique de la Langue (NLP), Machine Learning, Programmation (Python, Java, etc.), Mathématiques, Large Language Models
- Bachelor's DegreeUniversité Paris-Cité2020Licence en NLP: Traitement Automatique de la Langue (NLP), Machine Learning, Programmation (Python, Java, etc.), Mathématiques