Position Overview
Dice is the leading career destination for tech experts at every stage of their careers. Our client, Accion Labs, is seeking a Machine Learning Engineer (Multi-Modal Retrieval & Conversational AI) to design and implement AI systems that combine computer vision, NLP, and structured data for retrieval, similarity search, and conversational interactions. Based in Dallas, Texas (2 days from office and rest of the days from home), you will work on multi-modal pipelines that extract features, perform matching, and interact with end-users through schema-driven conversational flows.
Key Responsibilities
- Develop and optimize embedding pipelines for image and text similarity (e.g., CLIP, SigLIP, Sentence Transformers).
- Implement vector search and retrieval using FAISS, Pinecone, or pgvector.
- Build feature extraction pipelines (OCR + NER + numeric parsers) to detect schema-defined attributes.
- Design and validate a feature comparison engine to detect missing or low-confidence values.
- Integrate conversational AI agents with slot-filling logic to request missing details from users.
- Apply server-side validation for numeric, categorical, and free-text inputs.
- Track experiments using MLflow / W&B and evaluate with metrics like Recall@k, MRR, F1.
- Deploy retrieval and conversational services on Kubernetes / App Services with CI/CD pipelines.
- Collaborate cross-functionally with engineers and product teams to refine schema and conversational UX.
Required Qualifications
- Strong knowledge of embeddings and retrieval models for multi-modal data.
- Experience with OCR + NER pipelines for structured data extraction.
- Proficiency in vector databases (FAISS, Pinecone, pgvector, Azure AI Search).
- Familiarity with LLM integration for conversational AI (tooling, slot-filling, schema control).
- Strong Python and PyTorch/TensorFlow skills.
- Hands-on experience with containerized ML services (Docker, Kubernetes).