Ravik Daru
Ai Engineer Data Scientist @IQVIA
Signup · Get unlimited contacts
WORK HISTORY
Ai Engineer Data Scientist @IQVIA
GB
Designed and deployed Agentic AI systems using Python and Ray RLlib for autonomous decision-making and adaptability in enterprise workflows.Built RL pipelines in PyTorch and TensorFlow with DQN and PPO to optimize policy learning and agent efficiency.Modeled MDPs and POMDPs for probabilistic reasoning, improving decision-making under uncertainty.Developed multi-agent communication using gRPC and REST APIs for seamless coordination across distributed systems.Performed Multi-Agent RL (MADDPG, QMIX) for optimizing cooperative and competitive agent behaviors.Integrated OpenAI Gym and simulation setups for scalable agent training and testing.Used Unity ML-Agents and SimPy to analyze emergent behaviors and improve system stability.Implemented meta-learning with MAML in JAX for rapid adaptation and better generalization.Automated Agentic AI workflows with n8n integrating LangChain agents, Hugging Face models, and enterprise APIs for real-time decision routing.Designed task planning with PDDL and Fast Downward Planner for multi-step execution.Automated multi-agent training workflows via Airflow and Prefect, improving reproducibility.Deployed AI ecosystems with Docker and Kubernetes for scalable, production-ready systems.Developed evaluation frameworks using NumPy, Pandas, and Matplotlib for data-driven reporting.Analyzed agent interactions with NetworkX to identify optimization opportunities.Implemented continual learning via self-play and curriculum learning for progressive improvement.Performed reward shaping and ethical constraints to ensure reliable, safe AI behavior.Enabled natural language interfaces using OpenAI GPT APIs and Hugging Face Transformers to combine Agentic and Generative AI.Integrated autonomous AI with GenAI extensions for scalable and robust platforms.Researched next-gen Agentic AI using transformer-based decision models and LangChain with LLaMA for enhanced reasoning.
ABOUT RAVIK DARU
Over 13 years of IT industry experience delivering scalable AI/ML solutions with expertise in Agentic AI, Generative AI, MLOps, Deep Learning, and Data Science across multiple industries.Proficient in the end-to-end AI/ML ecosystem including Python, PyTorch, TensorFlow, JAX, Ray RLlib, n8n, Hugging Face, LangChain, Google Colab, Docker, Kubernetes, MLflow, Airflow, Spark, Kafka, Tableau, Power BI, Grafana, SQL, R, and SAS.Expert in Agentic AI frameworks such as Ray RLlib, Unity ML-Agents, OpenAI Gym, and LangGraph, enabling autonomous workflow orchestration and process optimization.Skilled in reinforcement learning (DQN, PPO) for adaptive decision-making and multi-agent RL (MADDPG, QMIX) for collaborative and competitive behavior in enterprise workflows.Experienced in multi-agent communication using gRPC and REST APIs for synchronized execution and integration with IT systems.Proficient in integrating Agentic AI with n8n workflow automation, enabling low-code orchestration of multi-agent pipelines, API calls, and LLM-driven decision automation.Applied MDPs and POMDPs for probabilistic reasoning, improving predictive and decision-support reliability.Expert in simulation-driven AI training using Unity ML-Agents, OpenAI Gym, and SimPy.Implemented meta-learning and continual learning including MAML (JAX), self-play, and curriculum learning to enable adaptability and performance improvement.Skilled in task planning pipelines with PDDL and Fast Downward Planner for automated multi-step execution.Experienced with Generative AI LLM frameworks including Hugging Face Transformers, OpenAI GPT APIs, LLaMA, and LangChain for context-aware reasoning and RAG workflows.Developed transformer-based architectures in PyTorch and TensorFlow for text generation, summarization, and document intelligence.Fine-tuned LLMs using LoRA, PEFT, and transfer learning.Designed RAG pipelines with LangChain, FAISS, Pinecone, and Weaviate, applying prompt engineering (temperature, top-p) for controllable outputs.Expert in next-generation GenAI methods including diffusion models and RLHF, implementing reward shaping, moderation, and explainability frameworks to ensure ethical AI.Optimized inference with ONNX Runtime, TorchScript, quantization, pruning, and mixed-precision training.Experienced in deep learning (CNNs, RNNs, LSTMs, Autoencoders, GANs), computer vision (YOLO, Faster R-CNN, SSD, U-Net, SegNet), classical ML/statistics (Scikit-learn, R, SAS), and enterprise visualization (Tableau, QlikView, Dash, Matplotlib, ggplot2).
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.