Deep Learning Engineer
CodeRound is hiring for this role.
- Deep Learning
- Machine Learning
- Artificial Intelligence
- LLM
- VLM
- Foundation Models
- Transformer Architectures
- PyTorch
- TensorFlow
- JAX
- NLP
- Computer Vision
- Multimodal AI
- OCR
- Document AI
- Information Extraction
- Information Retrieval
- LayoutLM
- Vision Transformers
- Model Training
- Fine-Tuning
- SFT
- PEFT
- LoRA
- QLoRA
- RLHF
- RLAIF
- Representation Learning
- Distributed Training
- GPU Optimization
- Model Evaluation
- Model Serving
- vLLM
- TensorRT
- Triton
- SGLang
- MLOps
- Production ML
- Python
We're hiring a Deep Learning Engineer to build and deploy state-of-the-art deep learning architectures that solve complex real-world problems involving unstructured data, document intelligence, and multimodal understanding. You'll work on training and fine-tuning foundation models, developing generalized architectures, and pushing the boundaries of NLP, computer vision, LLMs, and VLMs. This role is ideal for engineers who enjoy going beyond API integrations to build, optimize, and deploy production-grade models that deliver measurable improvements in accuracy, performance, and scalability.
What you'll do
- Build State-of-the-Art Models: Design, train, and deploy advanced deep learning architectures to solve complex business problems involving unstructured data and intelligent information extraction.
- Train & Fine-Tune Foundation Models: Lead hands-on experimentation with LLMs, VLMs, and specialized deep learning models using pre-training, fine-tuning, and parameter-efficient adaptation techniques.
- Develop Generalized Architectures: Build reusable models that generalize across datasets and tasks, reducing dependence on manual feature engineering and task-specific model customization.
- Advance Document Intelligence: Develop high-performance models for OCR, document classification, hierarchical information extraction, and structured extraction from complex documents and tables.
- Experiment with Emerging Techniques: Research and implement advancements in Transformer architectures, multimodal learning, representation learning, and model optimization to improve model capabilities.
- Drive Model Evaluation: Build robust evaluation frameworks, establish benchmarks, analyze failure cases, and systematically improve model accuracy, generalization, and reliability.
- Optimize Training & Inference: Improve training efficiency, GPU utilization, inference throughput, latency, and computational costs across the model lifecycle.
- Deploy Models at Scale: Take models from experimentation to production by developing scalable serving pipelines, integrating ML infrastructure, and ensuring reliable performance under real-world workloads.
- Own End-to-End Model Development: Drive the complete model development lifecycle, from dataset preparation and architecture selection to training, evaluation, deployment, and continuous improvement.
Must have
Deep Learning Experience: 4–10+ years of experience in Machine Learning/Deep Learning, with at least 2–3 years of hands-on experience training or fine-tuning deep learning models. Hands-on Model Training: Demonstrated experience with pre-training, continued pre-training, supervised fine-tuning (SFT), parameter-efficient fine-tuning (PEFT), LoRA/QLoRA, knowledge distillation, or domain adaptation. Foundation Models & Transformers: Strong understanding of Transformer architectures, attention mechanisms, and modern foundation models, including LLMs and Vision-Language Models (VLMs). Specialized Deep Learning Expertise: Strong hands-on experience in at least one domain, such as NLP, computer vision, OCR/document AI, multimodal learning, information retrieval, or speech/audio processing. Deep Learning Frameworks: Proficiency in PyTorch, TensorFlow, or JAX, with strong fundamentals in model architecture, representation learning, optimization, and evaluation. Model Evaluation & Optimization: Experience designing evaluation pipelines, benchmarking models, analyzing datasets, and improving metrics such as accuracy, F1 score, recall, precision, or word error rate. Production-Grade ML Systems: Proven experience deploying deep learning models into production, with an understanding of model serving, inference optimization, scalability, and reliability. Strong Software Engineering: Proficiency in Python, version control, testing, CI/CD, and writing maintainable, production-quality code. Experimental Mindset: Ability to independently investigate emerging research, rapidly experiment with new architectures and training techniques, and translate promising approaches into practical solutions.
Good to have
Modern LLMs & VLMs: Experience working with models such as Llama, Qwen, Mistral, Gemma, LLaVA, or other open-source and proprietary foundation models. Document AI & OCR: Experience building models for document understanding, OCR, document classification, key information extraction, and structured data extraction from invoices, receipts, identity documents, and other complex documents. Advanced Document Understanding: Familiarity with layout-aware Transformers such as LayoutLM, hierarchical information extraction, complex table parsing, and spatial relationships within documents. Computer Vision Architectures: Experience with Vision Transformers (ViTs), CNNs, image understanding, image preprocessing, or multimodal vision-language architectures. Advanced Training Techniques: Exposure to self-supervised learning, reinforcement learning from human or AI feedback (RLHF/RLAIF), synthetic data generation, and advanced fine-tuning strategies. Distributed Training & Infrastructure: Experience with multi-GPU training, distributed computing, large-scale datasets, and ML infrastructure for training and experimentation. Inference Optimization: Familiarity with vLLM, TensorRT, Triton, or SGLang, along with quantization, pruning, distillation, and high-throughput inference optimization. MLOps & Deployment: Experience with Kubernetes, model monitoring, automated evaluation pipelines, and production ML infrastructure across cloud or on-premises environments. Research & Benchmarking: Contributions to research publications, open-source deep learning projects, novel model architectures, or measurable improvements over existing state-of-the-art benchmarks.