Research-driven AI Engineer with 3+ years of experience building and deploying deep-learning systems across multimodal perception, generative models, and large-scale training pipelines. Specialized in training policies using diverse trajectory data and operating complete data-to-deployment loops.
Most AI models in production don't fail in isolation; they fail due to dependency failures. This session explored why well-trained models silently degrade in the real world, covering practical approaches to monitoring, evaluation, drift detection, guardrails, and operational reliability for production AI deployments. Built around the through-line: "A model is a photograph of a moving world."
Audience Delegates, AI Leaders, Industry practitioners, researchers, and AI product teams
Developed a novel dual-encoder architecture for multimodal synthesis (Sketch + Image), focusing on interpretability and fine-grained control over generated outputs.
Prototyped a real-time Sign Language Generation proof-of-concept within 48 hours using MediaPipe and pose estimation models for gesture-to-avatar translation.
Built a gender classification model using PCA and SVM, optimizing feature extraction pipelines for high accuracy.
Authored technical analysis of Generative AI and Virtual Try-On Systems, covering model architectures, pipelines, and industry trends.
PyTorch, TensorFlow, Transformers, Diffusion Models, Neural Networks, CLIP, StyleGAN, Model Training, Hyperparameter Tuning, Algorithm Development
Docker, Kubernetes, CI/CD, GPUs (CUDA), GCP, AWS (S3, Lambda), Azure Compute, FastAPI, Flask, Distributed Training
OpenCV, Semantic Segmentation, Vision-Language Models, Video & Trajectory Data, Multimodal Learning, Image Processing
Python, C++, Bash, SQL, NumPy, Pandas, Matplotlib, Scikit-learn, Data Pipelines, Synthetic Data Generation