Brooklyn, New York, United States
AI and Computer Vision Expert | Transforming Broadcast Analytics with Cutting-Edge Technologies Innovative Machine Learning Leader with 7+ years of experience driving AI solutions at Nielsen. Specializing in Large Language Models, Computer Vision, and MLOps, I architect sophisticated AI pipelines that revolutionize media content analysis. Key Expertise: Advanced AI/ML Technologies Large Language Models Computer Vision Cloud-Native Solutions Passionate about leveraging AI to unlock actionable insights and drive technological innovation.
Leading innovation in agent-based systems and AI workflows. Building enterprise AI platforms for marketers and analysts, designing serverless multi-cloud infrastructure for document processing and LLM orchestration, developing multi-hop RAG systems for financial analytics, implementing rigorous RAG evaluation with independent judges, and creating cultural intelligence pipelines that turn social data into brand insights and creative direction. Stack: Python, LanceDB, Pinecone, Claude + GPT +Gemini, Hindsight, Agno.
As AI Lead, I spearhead a team of 5 Machine Learning Engineers, orchestrating the development and deployment of cutting-edge AI solutions for broadcast analysis. Leading the architecture of the Analysis Tribe, I've implemented sophisticated AI pipelines utilizing AWS Bedrock for LLM inference (Claude3.5-sonnet), vLLM for distributed serving of Llama-3.3 70B, and NVIDIA Triton for computer vision models. Our stack leverages state-of-the-art technologies including Ultralytics YOLOv11, SAM2 for object tracking, and various object detection frameworks (TF-Models, TorchVision). I've architected solutions combining LangChain for vector embeddings with MongoDB, and optimized model performance using ONNX and TensorRT. These implementations have significantly enhanced TV ratings predictions, automated sports event scope analysis across multiple platforms, and enabled real-time detection of brands, celebrities, and highlights in broadcasts through computer vision and audio recognition. My role encompasses full-stack development, DevOps practices, and algorithmic engineering, driving innovation in broadcast analytics through AI-powered solutions.
As the team leader of four ML developers, my role involves guiding and supporting them to achieve optimal results in deep learning. I am also the lead architect of the analysis tribe, which is the primary focus of the Nielsen Sport division. Two major projects that I oversee include exposure analysis and video segments. In the video segments project, I run a complex pipeline that incorporates over 1000 models. Each video stream has its own segmentation, which helps us to better understand the brand exposure analysis pipeline that identifies logos in sports events. Our deep learning training is based on PyTorch and Tensorflow, which run on the Sagemaker pipeline. For serving, we rely on Nvidia Triton, which runs on K8s with Argoflow or Step Function. Our data is hosted on Mongo and Redshift, and we use AWS as our primary cloud provider. As the core business, we face many challenges, such as running more than 10k models on a single endpoint, executing a pipeline of over 10 different models, and collecting results that will lead to better outcomes. Additionally, we work with strict SLAs to ensure that video data segmentation is completed within minutes of arrival and exposure to the client.
I am the lead computer vision engineer in Vbrand, which Nielsen acquired. My daily tasks are to implement deep learning applications using cutting-edge technology for object detection, semantic segmentation, and video action recognition. These tasks are built with Tensorflow or PyTorch and deploy to the Docker environment that runs in the Kubernetes cluster at AWS. Python is my primary programming language, and I developed an infrastructure that supports video analysis, data clustering, and complex streaming pipelines. The infrastructure developed optimization methods such as stream processing (Kafka, RabbitMQ) and distributed computing (Dask, Spark).
Programming experience in C#,Python. Vast knowledge in the big data eco-system: Mongo, Spark ,ElasticSearch. Created clusters from scratch. Experience in coding ETL processes, creating database schemas and infrastructure, and of NoSQL databases. Hands on experience in building and operating big data infrastructures from the selection of the right mix of technologies, setting up the cluster servers, designing and optimising schemes, and tuning performance. Develop and deploy massive real time data processing infrastructures in AWS. Design and build machine learning algorithms. Support and develop large amounts of data, keeping scalability in mind. Take part in all development stages – from design to deployment
Developed new user-facing features. Build reusable code and libraries for future use. Ensured the technical feasibility of UI/UX designs. Optimised application for maximum speed and scalability. Assured that all user input is validated before submitting to back-end. Collaborated with other team members. Proficient understanding of web markup, including HTML5, CSS3 Basic understanding of server-side CSS pre-processing platforms, such as LESS and SASS. Proficient understanding of client-side scripting and JavaScript frameworks, including jQuery. Good understanding of JavaScript libraries and frameworks, such as AngularJS. Good understanding of asynchronous request handling, partial page updates, and AJAX. Good knowledge of image authoring tools, to be able to crop, resize, or perform small adjustments on an image. Excellent understanding of cross-browser compatibility issues and ways to work around them. Excellent understanding of code versioning tools, such as Git & SVN Good understanding of SEO principles and ensuring that application will adhere to them.