Welcome to IJW!

Why IJW?

We have spent most of our careers building tools for people to get things done faster, cheaper and more easily. Those tools just worked, so why not call our new startup something simple and easy to remember?

In this crazy era of unprecedented growth of new technologies, in particular Generative AI, we are building new tools - that just work - for developers to solve complex problems.

"Anyone that's used AI tools over the last year knows that they're extremely useful for simple problems but as soon as you get into more complex problems they become counterproductive."

Jeff Delaney, Fireship (YouTuber) 13 March, 2024

What We Do

We're developing cutting-edge, never-before-seen AI solutions using the latest advancements in GPTs, open-source SOTA models, and a whole lot of secret sauce. Our technology is poised to revolutionize how businesses interact with and leverage content and its data.

Check out StarZero (*0), our groundbreaking video search platform that lets you find any moment in any video using natural language.

Our Focus Areas

Democratizing Generative AI:

We're building developer-friendly solutions based on Open-Source SOTA models that are easy to integrate, scale, and won't lock you into a specific vendor.

Video Understanding:

With our flagship product StarZero (*0), we're revolutionizing video search. Find any moment, inside any video, instantly. Our technology combines real-time video understanding with natural language search, going beyond traditional video platforms to unlock the full potential of your video library.

Advanced Embeddings:

We offer a range of powerful embedding solutions, from text and visual embeddings to multimodal representations, allowing you to uncover hidden connections and insights within your content and data.

Ready to unlock the future of content understanding? Contact us today!

Technologies

We leverage a powerful stack of AI technologies, including:

• LLMs: Using SOTA Open-Source LLMs to power content generation with advanced natural language understanding.
• VLMs (Video-Language Models): Video understanding with Large Language Models. Answering questions, summarizing, analyzing.
• Clustering: Grouping data points to uncover natural divisions within your datasets
• Semantic Search: Enhancing search capabilities with context-awareness
• NLU (Natural Language Understanding): Extracting valuable insights from text with Entity Recognition, Entity Typing, Linking and Relationship Extraction.
• Knowledge Graphs: Mapping complex relationships in data to facilitate deeper analysis and understanding
• Graph RAGs: Enhancing RAG with knowledge graphs, hierarchies, and summaries for improved reasoning and task performance
• Automated Data Cleansing: Enhancing data quality through automated detection and correction of errors and inconsistencies.
• Deep Analytics: Offering deep data insights to uncover patterns
• Video Understanding: Analyzing and understanding video content using advanced AI techniques
• Video Retrieval: Efficiently searching and retrieving relevant video content based on visual and semantic features
• Visual Models: Implementing state-of-the-art visual models for image and video analysis
• Face Recognition: Detecting and recognizing faces in images and videos with high accuracy
• Speaker Recognition: Identifying and distinguishing speakers in audio and video content
• Object Tracking: Object Detection + Tracking + reID + Classification + Segmentation
• Embeddings
• Text Embeddings and Similarity: Transforming text into numerical vectors to discover hidden similarities and patterns.
• Visual Embeddings: Discrete Image embeddings transforming images and frames into features for classification or to uncover hidden similarities and patterns.
• Visual Retrieval with OpenCLIP: Contrastive Language-Image embeddings to match and retrieve visual elements via text descriptions.
• Zero-shot Text to Audio Retrieval: Latent representations of any audio and text for text to audio retrieval.
• Face Recognition and Face Embeddings: Latent representations of any faces for face recognition.
• Speaker Diarization and Speaker Embeddings: Speaker Diarization and Latent representation of speakers for multi-audio recognition and linking.
• Multimodal embeddings: Multimodal representations to bind multiple modalities into a single space.