AI/ML Engineer
Specializing in advanced machine learning, artificial intelligence, and MLOps, expertise arge Language Models (LLMs),
Optical Character Recognition (OCR), Reinforcement Learning (RL) ML pipelines.
Technical Proficiencies:
Large Language Models (LLMs):
• transformer architectures (GPT-3.5, GPT-4, BERT, T5)
• attention mechanisms positional encodings
• Advanced prompt engineering: chain-of-thought, few-shot learning
• Efficient inference: quantization, pruning, knowledge distillation
• Tokenization: BPE, WordPiece, SentencePiece
• Contrastive learning self-supervised pretraining
• Parameter-efficient fine-tuning LoRA, prefix tuning, adapters
• Used Azure OpenAI, Amazon SageMaker, Google Vertex AI
Optical Character Recognition (OCR):
• End-to-end OCR pipelines using CNNs and RNNs
• Scene text detection: EAST, TextBoxes++, CRAFT
• CTC loss attention sequence-to-sequence recognition
• Post-processing: lexicon correction, N-gram language models
• Document layout analysis with GNNs and transformers
• Image preprocessing: binarization, skew correction
• Handwritten text recognition HMMs neural networks
• Integrated Azure, AWS, Google Cloud OCR services
Reinforcement Learning:
• Value-based (DQN) policy-based (TRPO, PPO, SAC) algorithms
• Custom environments OpenAI Gym Unity ML-Agents
• Advanced exploration: intrinsic motivation, curiosity-driven
• Multi-agent RL: MADDPG, QMIX, MAPPO
• Inverse RL: MaxEnt IRL, GAIL
• Off-policy learning experience replay optimization
• Hierarchical RL complex tasks
• Azure Bonsai, AWS DeepRacer, Google AI Platform
MLOps and Infrastructure:
• Model versioning: MLflow, DVC, cloud-specific registries
• Experiment tracking: W&B, Neptune.ai, cloud solutions
• Feature stores: Feast, Tecton, cloud-specific offerings
• Data versioning: Delta Lake, Pachyderm, cloud services
• A/B testing: Optimizely, Google Optimize, Azure Experimentation
• Distributed training: Horovod, DeepSpeed, cloud solutions
• Scalable inference: TF Serving, Triton, cloud services
• Edge deployment: TF Lite, ONNX Runtime, TensorRT, IoT solutions
Additional ML/AI Expertise:
• PyTorch TensorFlow: custom layers, distributed training
• Neural architecture search: DARTS, ENAS
• Federated learning differential privacy
• Custom loss functions data pipelines
• Explainable AI: SHAP, integrated gradients
• AutoML: Optuna, Ray Tune, cloud-specific solutions
• Advanced ensembles: stacking, blending, XGBoost, LightGBM
Open-source Contributions:
• Hugging Face Transformers: custom attention mechanisms
• OCR packages for tesseract and OpenCV
• Custom RL environments for OpenAI Gym
• MLflow: support for new ML frameworks
• Feast feature
algorithm analysis, distributed systems, HPC. Committed scalable, ethical AI solutions
robustness production-ready ML systems cloud platforms open-source.