Tensorway

Our Data Extraction solution streamlines critical financial operations:

• Automates the extraction and structuring of data from financial statements for quick analysis, aiding in faster, informed decision-making.
• Enhances invoice management by automatically extracting key information and initiating payments, reducing manual effort and increasing accuracy.
• Streamlines the loan approval process by efficiently analyzing financial documents to assess creditworthiness, minimizing errors and biases.
• Facilitates AML and KYC checks by automating data extraction and verification from customer documents, improving compliance accuracy and efficiency.

Using NLP, Tensorway created an advanced system able to give students detailed human-like feedback on their essays.

The AI tutor allowed our client to process more student work at a time, making GAMSAT preparation more efficient. Thanks to the AI tutor, the client could focus on other educational projects and more guided mentoring of his students.

Tensorway’s model provided detailed feedback on various essay metrics such as overall quality, structure, clarity, and specific sections. This helped students understand their strengths and areas for improvement more clearly, ultimately enhancing their essay-writing skills. Notably, the feedback was so human-like that the client was amazed at how closely it resembled his own writing style.

The solution allows students not just to receive feedback on one essay, but iteratively improve the essay and evaluate it automatically, keeping the tutor away from the process. It is also cheaper and makes education more accessible.

The client is a Computer Vision-powered instant ID verification service for fraud detection and Know Your Customer (KYC).

Our team:

• Developed a model to detect unreadable documents
• Worked on different techniques to recognize glare-obscured documents
• Boosted model performance by enhancing the dataset using synthetic documents
• Adapted heavy models to run effectively even on mobile devices

Problem statement

A bot to answer questions related to medical or psychological field

Our solution

Trained a Large Language Model on dialogues with clients and provided it with medical materials as context. During dialogue, model finds most relevant materials and uses them for an accurate answer

Impact

Reduced load of staff, increased customer satisfaction

Kwizie is a pioneering tool for enhancing video-based learning embraced by educators, learners, and content creators. It allows users to create engaging quiz games from videos of various lengths, including extensive MOOC lectures and brief ad-hoc tutorials. This tool stands out for its fast, intuitive interface and data-driven approach to learning, offering powerful analytics to refine educational strategies.

• Fast and intuitive quiz game creation from videos
• Data analytics for optimized learning strategies
• Multiplayer functionality for collaborative learning
• Compatibility with videos up to 3 hours long
• Integration with YouTube for diverse video sources
• Easy sharing options via QR code or link
• Mobile-friendly and device-agnostic platform

A solution to help newspaper publishers better know their audience. Tensorway built an ML model at scale to segment audience into explainable "personas" using only user activity on a website - no personal data involved. Then, we utilized behavioral analytics to help publishers expand audience and determine the content to prioritize, following the extraction of lots of insights using data analysis and deploying the model to production.

Problem statement

Given a prompt (year, place, event), generate historical map

Existing image generation solutions

Trained on diverse dataset, does not know history

Our solution

Can leverage historical information, can generate maps in needed format

Problem statement

Generate high-quality interior 3D model using smartphone only

Existing approaches "weakness"

Detecting scene as a single low-quality object with inseparable smaller objects

Our solution

Detect each object separately so that they're instantly movable and replaceable in a game engine

A Machine Learning model that generates accurate text descriptions of what's inside the room based on the room's photo.

This allowed the client to automate image description generation and achieve an extreme level of detail: not only the model identifies placement of objects and surfaces, but also their color, material, and texture, as well as distinguishes smaller objects that older models used to neglect.