Gen AI Engineers (Claude code, Codex, Cursor)

Accellor is an AI-native services firm purpose-built for the post-ChatGPT era. Free from legacy constraints, we focus on delivering measurable business outcomes through advanced AI, data, and engineering capabilities. Our mission is to operationalize AI at scale and unlock sustained enterprise value. Our offerings span AI solutions, data services, enterprise applications, and product engineering, tailored to industry-specific needs across healthcare, life sciences, telecom, retail, financial services, and technology. By leveraging design thinking and technology-agnostic architectures, we ensure faster time-to-value and seamless interoperability. With a proven track record of enabling Fortune 100 enterprises and global innovators, Accellor stands as a trusted partner for organizations seeking to harness the full potential of AI. Our vision is clear: to build intelligent, connected ecosystems that deliver measurable outcomes and redefine the future of enterprise transformation. Role Summary: We are looking for Gen AI Engineer with strong problem-solving ability, exceptional Claude code, codex expert, cursor, solid Python engineering skills, and hands-on understanding of machine learning, deep learning, GPUs, and CUDA fundamentals. The ideal candidate may be early in their career but must be technically sharp, curious, implementation-driven, and capable of learning complex AI systems quickly. This role is suited for someone who can build models, debug training issues, optimize GPU workloads, understand tensor operations, and work closely with research, platform, and engineering teams. Key Responsibilities: Design, build, train, and evaluate AI/ML models using Python, TensorFlow, PyTorch, or JAXSolid skills with Claude code, codex, CursorCan do large number of PRs per monthDevelop clean training pipelines with data loading, checkpointing, logging, validation, and experiment trackingWork on deep learning models including CNNs, Transformers, LLMs, embeddings, and attention-based architecturesSolid builder with architect mindsetDebug model issues such as poor convergence, overfitting, unstable loss, NaNs, tensor-shape errors, and GPU memory failuresRun and optimize models on GPUs with awareness of CUDA execution, memory usage, batching, mixed precision, and kernel performanceProfile training and inference workloads to identify bottlenecks in compute, memory, data loading, and communicationSupport development of custom operators, fused kernels, CUDA/Triton-based optimizations, or framework-level performance improvementsBuild AI inference pipelines with focus on latency, throughput, reliability, cost, and qualityCreate evaluation pipelines for model accuracy, robustness, hallucination risk, safety, latency, and regression testingRead research papers, implement ideas, run experiments, and clearly document findings, trade-offs, and limitations Requirements Strong programming experience in PythonGood understanding of data structures, algorithms, numerical programming, and clean code practicesHands-on experience with TensorFlow, PyTorch, or JAXStrong fundamentals in: Deep learning Backpropagation Gradient descent Optimizers Loss functions Regularization CNNs Transformers Attention mechanisms Embeddings Model evaluationExperience training and debugging models on CUDA-enabled GPUsBasic understanding of: GPU memory CUDA kernels Threads, blocks, and warps Memory bandwidth Kernel launch overhead Mixed precision training CUDA out-of-memory debuggingComfortable with Linux, Git, Docker, shell scripting, and experiment managementAbility to reason from first principles and solve ambiguous technical problems Preferred Skills: CUDA C/C++ or Triton kernel developmentC++ exposure for performance-critical AI systemsDistributed training using PyTorch DDP, TensorFlow Distributed Strategy, JAX, NCCL, MPI, or similar toolsExperience with GPU profiling tools such as: NVIDIA Nsight Systems NVIDIA Nsight Compute PyTorch Profiler TensorFlow ProfilerExperience with inference optimization, quantization, TensorRT, ONNX, XLA, or model compilationExposure to LLMs, RAG systems, agents, embeddings, multimodal AI, or AI evaluation frameworksUnderstanding of FP32, FP16, BF16, INT8, tensor cores, and numerical stability What Excellent Looks Like: Can implement a model from a research paper without blindly copying codeCan debug why a model is not learningCan explain why GPU utilization is lowCan identify whether a workload is compute-bound, memory-bound, or communication-boundCan write clean, reproducible, and testable Python codeCan profile before optimizingCan work across model logic, GPU execution, data pipelines, and inference systemsCan communicate technical findings clearly and precisely Ideal Candidate Profile: The ideal candidate is an early-career AI engineer who is not limited to using ML libraries as black boxes. They understand how models train, how tensors move, how GPUs execute workloads, and how performance, quality, reliability, and safety come together in real AI systemsThe strongest candidates will show hands-on projects in model implementation, GPU acceleration, CUDA/Triton operators, distributed training, LLM evaluation, inference optimization, or research paper reproduction