Auriga - Generative AI & Computer Vision Engineer

Description : Were seeking a hands-on GenAI & Computer Vision Engineer with 35 years of experience delivering production-grade AI solutions. You must be fluent in the core libraries, tools, and cloud services listed below, and able to own end-to-end model developmentfrom research and fine-tuning through deployment, monitoring, and iteration. In this role, youll tackle domain-specific challenges like LLM hallucinations, vector search scalability, real-time inference constraints, and concept drift in vision models.Key Responsibilities : Generative AI & LLM Engineering : - Fine-tune and evaluate LLMs (Hugging Face Transformers, Ollama, LLaMA) for specialized tasks. - Deploy high-throughput inference pipelines using vLLM or Triton Inference Server. - Design agent-based workflows with LangChain or LangGraph, integrating vector databases (Pinecone, Weaviate) for retrieval-augmented generation. - Build scalable inference APIs with FastAPI or Flask, managing batching, concurrency, and rate-limiting.Computer Vision Development : - Develop and optimize CV models (YOLOv8, Mask R-CNN, ResNet, EfficientNet, ByteTrack) for detection, segmentation, classification, and tracking. - Implement real-time pipelines using NVIDIA DeepStream or OpenCV (cv2); optimize with TensorRT or ONNX Runtime for edge and cloud deployments. - Handle data challengesaugmentation, domain adaptation, semi-supervised learningand mitigate model drift in production.MLOps & Deployment : - Containerize models and services with Docker; orchestrate with Kubernetes (KServe) or AWS SageMaker Pipelines. - Implement CI/CD for model/version management (MLflow, DVC), automated testing, and performance monitoring (Prometheus + Grafana). - Manage scalability and cost by leveraging cloud autoscaling on AWS (EC2/EKS), GCP (Vertex AI), or Azure ML (AKS).Cross-Functional Collaboration : - Define SLAs for latency, accuracy, and throughput alongside product and DevOps teams. - Evangelize best practices in prompt engineering, model governance, data privacy, and interpretability. - Mentor junior engineers on reproducible research, code reviews, and end-to-end AI delivery.Required Qualifications : You must be proficient in at least one tool from each category below : .LLM Frameworks & Tooling : - Hugging Face Transformers, Ollama, vLLM, or LLaMA.Agent & Retrieval Tools : - LangChain or LangGraph; RAG with Pinecone, Weaviate, or Milvus.Inference Serving : - Triton Inference Server; FastAPI or Flask.Computer Vision Frameworks & Libraries : - PyTorch or TensorFlow; OpenCV (cv2) or NVIDIA DeepStream.Model Optimization : - TensorRT; ONNX Runtime; Torch-TensorRT.MLOps & Versioning : - Docker and Kubernetes (KServe, SageMaker); MLflow or DVC.Monitoring & Observability : - Prometheus; Grafana.Cloud Platforms : - AWS (SageMaker, EC2/EKS) or GCP (Vertex AI, AI Platform) or Azure ML (AKS, ML Studio).Programming Languages : - Python (required); C++ or Go (preferred).Additionally : - Bachelors or Masters in Computer Science, Electrical Engineering, AI/ML, or a related field. - 35 years of professional experience shipping both generative and vision-based AI models in production. - Strong problem-solving mindset; ability to debug issues like LLM drift, vector index staleness, and model degradation. - Excellent verbal and written communication skills.Typical Domain Challenges Youll Solve : - LLM Hallucination & Safety : Implement grounding, filtering, and classifier layers to reduce false or unsafe outputs. - Vector DB Scaling : Maintain low-latency, high-throughput similarity search as embeddings grow to millions. - Inference Latency : Balance batch sizing and concurrency to meet real-time SLAs on cloud and edge hardware. - Concept & Data Drift : Automate drift detection and retraining triggers in vision and language pipelines. - Multi-Modal Coordination : Seamlessly orchestrate data flow between vision models and LLM agents in complex workflows. (ref:hirist.tech)