Specialized small language model development using AL/NLP - ON-1145

Genre de projet: Innovation
Discipline(s) souhaitée(s): Génie - informatique / électrique, Génie, Informatique, Sciences mathématiques
Entreprise: eXalt Solutions
Durée du projet: 4 à 6 mois
Date souhaitée de début: Dès que possible
Langue exigée: Anglais
Emplacement(s): Toronto, ON, Canada
Nombre de postes: 1
Niveau de scolarité désiré: MaîtriseDoctorat
Ouvert aux candidatures de personnes inscrites à un établissement à l’extérieur du Canada: No

Au sujet de l’entreprise: 

eXalt Solutions is redefining the future of B2B sales through AI.

Our no-code Knowledge Work as a Service (KWaaS) platform empowers companies to build Knowledge Bots that act as Advisors, Analysts, and Administrators—capturing expertise and automating decision-making at scale.

By turning institutional know-how into intelligent workflows, eXalt Solutions helps leading technology brands accelerate sales, reduce complexity, and grow faster. With AI-powered precision and a codeless foundation, we’re transforming how organizations sell, collaborate, and scale knowledge across their partner ecosystems.

Veuillez décrire le projet.: 

We are seeking an experienced AI/NLP engineer to lead the development of a domain-specific small language model (SLM) optimized for our proprietary dataset and use case. This model will power intelligent contextual understanding, summarization, and reasoning within several narrowly defined knowledge domains.

In addition, we aim to build a general-purpose training and inference environment designed to support multiple small language models serving distinct domains where content evolves on a weekly or monthly basis.

The ideal candidate combines deep technical expertise in language model architecture, fine-tuning, and evaluation with a strong appreciation for domain specificity, data quality, and cost-efficient deployment. Must be able to translate complex technical concepts for diverse audiences and work effectively with team members across all levels of technical expertise.

You will:
• Design and train specialized small language models (e.g., <10B parameters) using domain-specific data.
• Evaluate trade-offs between fine-tuning existing LLMs vs. training smaller custom architectures.
• Develop and maintain data ingestion, preprocessing, and annotation pipelines (structured and unstructured).
• Work closely with domain experts to curate and label data for supervised fine-tuning (SFT) and instruction tuning.
• Implement RAG pipelines, embedding stores, and retrieval optimizations for hybrid reasoning.
• Build and maintain evaluation benchmarks (accuracy, coherence, factual consistency, latency, cost).
• Integrate the model into production environments via API endpoints or lightweight serving frameworks.

Expertise ou compétences exigées: 

• MS or Phd in Computer Science, Data Science, Machine Learning, or related field.
• Proven experience developing or fine-tuning transformer-based models (e.g., BERT, LLaMA, Falcon, Mistral, Gemma).
• Strong Python skills and familiarity with Hugging Face Transformers, PyTorch, TensorFlow, LangChain, or vLLM.
• Hands-on experience with RAG systems, vector databases (REDIS, Pinecone, etc.), and embedding models.
• Understanding of tokenization, model compression, quantization, and distillation techniques.
• Experience deploying models via API endpoints, containers, or managed ML platforms (SageMaker, Vertex AI, etc.).
• Excellent problem-solving skills and ability to collaborate with non-technical domain experts.

PREFERRED EXPERIENCE
• Understanding in building large-scale data and analytics platforms
• Experience with building data systems from scratch with significant user-facing performance requirements
• Deep understanding of modern data technologies
• Strong opinions on data architecture backed by real-world experience at scale
• Expertise in backend API development (REST, GraphQL, or other modern approaches)
• Experience integrating AI/LLM technologies with structured data systems
• Excellent communication skills and ability to work closely with non AI experts
• Experience with governance, privacy, and bias-mitigation for proprietary models (desired but not required)