How to compete with Scale, Snorkel, and Appen without VC’s in 2023
Dedicated to our beloved advisor and a dear friend — Vlad.
Introduction
Hey, I am Theo. I work at PieData ai — a data annotation platform for Generative AI.
PieData in numbers: we used to have 50,000 human annotators, annotating more than a hundred million objects, and serving 400+ clients over the last 4 years.
7X increase in the demand for pre-annotated data
Interestingly, the dynamics of the market started shifting last year. We observed a whopping 7X increase in the demand for pre-annotated datasets. Such a paradigm shift not only signified the growing importance of ready-to-use data but also hinted at a potential revenue opportunity we were missing out on.
Snorkel was one of the first companies to offer a programmatic approach to data labeling. They grew in 4 short years from a promising seed start-up to a monster unicorn with a billion $ valuation.
What makes pre-annotated datasets so sought after?
Simply put, it’s the convenience and instantaneity. Traditionally, companies would engage in prolonged discussions with data annotation firms — from specifications to NDAs, data transfer to quality control, the process was intricate.
But imagine bypassing these steps and directly accessing a dataset tailor-made for a specific task. That’s the future we envisioned and set forth to achieve.
In our initial years, our strategy was direct and simple. We provided data scientists with free datasets tailored to their tasks.
The catch?
We retained all rights to this data, which enabled us to build a colossal data repository. Fast forward two years over 2000 data scientists collaborated with us, contributing to our 1 billion image data lake.
Today, with simple filters similar to LinkedIn, our clients can seamlessly access the datasets they need, ensuring their AI models are fed with the correct data.
Integrating PieDataLake with Manot to test your model performance before deployment
A few weeks ago George visited Yerevan and met with wonderful Ashot & Vazgen to talk about strong ML teams in Armenia that he should watch closely.
He mentioned Chinar founder of Manot and while we primarily deal with dataset generation Manot contributes by examining model performance to identify specific areas where the model might falter.
Their platform is crafted to serve both product teams and engineers, bridging the gap between these two roles that can frequently result in prolonged and inefficient feedback loops.
Manot studies the model’s weaknesses and comprehends its behavior across different scenarios that might be encountered in production. Based on this analysis, Manot provides insights in the form of images from the DataLake.
Check out PieData & Manot integration on Collab.
Through this integration, AI teams can accelerate the processes of model and data curation, and reduce costs in the machine learning lifecycle by approximately 50%. Recently, they published a case study on last-mile delivery. They have achieved remarkable results.
With 120 million photos spanning across categories like Polygons, Boxes, and Tags, the possibilities are immense.
We’re super thrilled to work closely with Manot’s team!
Chinar Movsisyan, the Founder and CEO, earned her master’s degree in computer science from Grenoble INP Esisar. She then pursued her Ph.D. under the mentorship of renowned computer vision scientist Sos Agaian.
CTO Konstantin Sargsyan, with over 15 years in software engineering, previously served as the principal technical solutions engineer at Vineti. There, he developed products for major pharmaceutical firms such as Johnson & Johnson and Takeda.
Heading up Manot’s research and development is Erik Harutyunyan. Erik earned his master’s degree in mathematics and data science from the Technical University of Munich, focusing on research in active learning for video data. He has over five years of experience as a machine learning engineer at SuperAnnotate.
The Team
I want to first mention our beloved angels who believed in our vision from the start.
Slava & Dmitry — Founders of Mighty Building (YC W17) & Intently.ai (YC W23)
Vitaly — ex. Founder of FoodFox and his ex. Co-founder — Anton
Folks, thank you for showing us your continued support. There is nothing better than having successful & experienced entrepreneurs on your side.
George Kaspar, our Co-founder & CTO, is unbelievable in the data science realm. His journey commenced with the photo editor ‘Teleport’ — an app that allowed users to transform their appearance through filters. They quickly amassed 1 million users in a week and caught the eye of Snap Inc., culminating in an acquisition worth $8M. He was 19 at the time.
George enhanced his expertise at PicsArt, devising state-of-the-art filters.
His achievements did not go unnoticed, with the former CEO of Teleport inviting him to head the AI team at Gradient, which amassed 150M active users in 3 years since launch.
A Few Words from Vlad, Advisor at PieData before THE END…
Having closely advised and observed the inner workings of PieData since January, I feel compelled to share a few insights that genuinely stand out.
Customer-Centric Learning: PieData isn’t just about building technology; it’s about understanding its users. Their consistent effort to glean knowledge from customer interactions is a testament to their dedication.
- Rapid Iterations: In the tech landscape, adaptability is key. The speed at which PieData iterates is not just impressive, it’s crucial for staying ahead.
- Lean Approach: I’ve seen many startups burn through capital, chasing the wrong goals. Not here. PieData’s ethos is built on scrappiness. Every dollar saved, every free solution explored, amplifies their commitment to sustainable growth.
- Rolling Up Their Sleeves: The team doesn’t shy away from getting their hands dirty. They’ll dive into the trenches, embrace unscalable solutions temporarily, all if it means gaining a clearer perspective on customer pain points and potential optimizations.
It’s this combination of attributes that makes me proud to advise and be a part of PieData’s journey.
