TrueFoundry: 2022 year-end Review

Dec 31, 2022: It’s amazing how introspective & excited ❣ I am feeling today! Maybe, because, year ends and beginnings have this special power that provide a reset & refresh button to our lives. However, the more I think about it the more I reflect that there is nothing fundamental about a year beginning and ending. It’s just an arbitrary man-made cycle and ideally, it is in our control to feel refreshed at the onset of a month, a week, a day or an hour — or anything in between for that matter! One learning I will try to carry in 2023!

With that set for 2023, I want to start by clapping 👏 & admiring 2022 for what an amazing year it has been for Machine Learning & MLOPs! There’s been an unprecedented advancement in AI in the last one decade but nothing has created the kind of excitement amongst common people as ChatGPT has- so much so, that Google declared code red! Besides, 2022 has seen Stable Diffusion from stability AI, Geoffrey Hinton’s feed forward training, AI generated matrix multiplication algorithms, 600M potential protein synthesis by Meta AI, bnb-int8 for most HuggingFace algorithms, Ray training 1M models in record time- such a phenomenal year!

On the other hand, in 2022, TrueFoundry has had such an impactful start- from hiring the first engineer on the team, to coding our rudimentary Streamlit based product UI, to hiring a world-class team, to building a multi-cloud Machine Learning deployment platform, to working with multiple $50B+ enterprises solving their production challenges! I am incredibly proud of our team for achieving such a huge feat in just 1 year! I would love to take a step back and reminisce ✍ on this beautiful first year for the TrueFoundry Team.

Team

What better way to start than talking about our team. In one year, we have grown to 15 technical and business people coming in from various backgrounds like Facebook, Amazon, Postman, Gojek & McKinsey. Our team is distributed across India, San Francisco & Paris with our first office-space in Bengaluru! We did our first company retreat this November which was one week of staying together, ideating, reflecting & playing!

Adoption

Although TrueFoundry is very early in its journey, we have been fortunate to co-build our platform with amazing design partners who are helping us with concrete actionable feedback on how different parts of our platform should be designed. Our work with them has allowed us to win paid proofs-of-concept with large enterprises and growth-stage startups. I am not ready to take their names just yet, but, we cherish the privilege of working with these companies!

Besides, we thank the startups and the developer community who have helped us with generous inputs by being early adopters of our product!

Product

Just a reminder for everyone, TrueFoundry is building a PaaS that enables developers to deploy & monitor their machine-learning models at the speed of big tech companies, cutting their production timelines from weeks to minutes. In one year, we have built-

• a cloud-agnostic platform on top of Kubernetes that allows people to deploy their models
• we have extended three user interfaces for deployments- python APIs for production, YAML configuration for CI/CD, GUI for experimentation
• intuitive infrastructure provisioning from one UI that makes the developers independent and allows the DevOps team to optimise cost better
• we have built terraform based one-click solution to deploy the entire platform on all 3 major cloud providers- AWS, GCP & Azure
• complete & automated logging and monitoring for your services
• model monitoring, data monitoring and drift-tracking dashboards
• complete experiment tracking with artefact management

Funding

TrueFoundry is proud and fortunate to have received the support and trust of marquee investors like Sequoia, Eniac & Naval Ravikant. Read a techcrunch article covering our seed funding announcement. Besides, we are fortunate to receive the support of angels and advisors like-

• Founder @ Kaggle (sold to Google — Anthony Goldbloom),
• Founder @ Rubrik (Soham Mazumdar)
• Builders Fund (Builders of ML platform at Salesforce),
• CTO @ Greenhouse (Mike Boufford)
• Chief Architect of FBLearner platform (Aditya Kalro)
• CTO/Founder @ Productiv (Ashish Aggarwal)
• Founders @ Whatfix (Khadim Bhatti & Vara Kumar)
• CIO @ Deutsche Bank (Dilip Khandelwal)
• CTO/Founder @ ClearCo (Satwik Seshasai)
• CTO @ AlphaSense (Raj Neervannan)
• Sr. ML Architect @ SnowFlake (Tal Shakhed) and
• Heads of ML at SentiLink, Migo, Upwork

Integrations

At TrueFoundry, we are building for the users and the goal is to meet the users where they are. For this reason, we analysed various tools, libraries, frameworks & platforms being used extensively and built out integrations with many of the popular ones-

• Machine Learning Libraries- tensorflow, pytorch, sklearn, xgboost, huggingface, lightgbm, mlflow etc.
• Model Serving Frameworks: TFServe, TorchServe
• Cloud Providers: AWS, GCP, Azure
• Code Versioning: GitHub, BitBucket
• CI/CD: Github Actions
• Docker Registries: DockerHub, AWS Container Registry, Google Container Registry
• Secret Management: AWS Parameter Store, Google Secrets Manager
• Dashboarding tools: Gradio, StreamLit
• Kubernetes

Events & Community

While we just got started on this end, some wonderful updates here include-

• Abhishek spoke in the 1729.world event by Fractal and Analytics Vidhya alongside speakers like Rajat Monga (Founder of TensorFlow), Sreenivas (Founder of Fractal). Here is the link to the Presentation.
• Nikunj presented at TechCrunch Disrupt 2022- startup BattleField
• Besides, we got an active slack community. Please join!
• Dozens of people from the community have contributed to TrueFoundry and the top projects are available here.
• Some of our most popular content include- Time killed my ML model, An impactful ML Model - how hard can it be?, Kubernetes for data scientists - Hosting predictions

Reflections

While reminiscing on this amazing year, we as founders, also took a step back to think and reflect on what went well and what didn’t go so well. We share our learnings, misses and achievements with you all-

Learnings

1. The traditional approach of working with design partners before the product is built doesn’t work well when you are building devtool that deal with production infrastructure. We spent the first 3–4 months doing this.
2. If an infra product takes X amount of time to build, its takes 2X the time to test and release. Underestimating the complexity was a mistake on our part that derailed the product release timelines.
3. Don’t design partner with companies that are far from your target market for ease of adoption. We spent a few months working with smaller companies while our product is intended for upmarket and enterprises. In retrospect, this was a rookie mistake.

Misses

We made a few mistakes in our execution that we need to course-correct:

1. We spent a lot of time talking to decision-makers within companies and getting their feedback on the product but ultimately the user of our platform are developers. We need to build a stronger channel to stay connected with developers!
2. As a company, we did not spend sufficient time evangelizing our platform, participating in conferences, giving talks and attending community events.
3. It’s understood that our product journey is long. We did not do a great job at breaking down into releasable milestones. Doing this would make our users stay more connected with the product.

Achievements

1. We built a very high quality team with great speed. Literally, every member of our team including the founders believe that this is the highest talent density team they have worked with, despite working with very high performing teams.
2. We are proud of the quality of the product and the speed of execution. We speak on this based on our personal experience and what we hear from our customers.
3. Excellent customer support and response time- almost every message from our customers have been responded to within a few minutes!

Year Ahead 2023

While 2022 has been a fun filled foundational year for TrueFoundry platform, 2023 is going to be even more exciting- as the platform gets delivered to more users and we prepare to get out of private beta. We welcome 2023 with the excitement of a kid that welcomes a new year in school. We are thrilled to innovate, build, learn and grow, once again in 2023 and create a positive impact in the world by making AI accessible to all.

Wishing you all a very Happy New Year! From Nikunj, Anuraag, Abhishek & the entire TrueFoundry team.

TrueFoundry is a ML Deployment PaaS over Kubernetes to speed up developer workflows while allowing them full flexibility in testing and deploying models while ensuring full security and control for the Infra team. Through our platform, we enable Machine learning Teams to deploy and monitor models in 15 minutes with 100% reliability, scalability, and the ability to roll back in seconds - allowing them to save cost and release Models to production faster, enabling real business value realisation.