Flipkart - Platform Architect - Machine Learning Models

MLP Team is responsible to ensure that at Flipkart , Data Scientists are able to train and host classic ML models, Large Language Models (LLM) and Large Image Models (LIM) at scale. This is a vastly exciting and rapidly evolving domain within Data Science.

As of July-2024, MLP Team has enabled Data Scientist to go Live with multiple use cases using in-house open source fine tuned models. Azure OpenAI ChatGPT 3.5, GPT 4 and GPT 4.o and Google Gemini 1.5 Pro and Flash models.

Products under MLP Team :

1. Modelhost : A python based framework allowing Data Scientists to host ML/DL models on K8s. Currently the framework is hosting 250+ models for online prediction hosting both CPU based models and GPU based large language and image models.

2. Runway : A platform to make GPU/CPU compute available for model development and training purposes.

3. Genvoy : A fully fledged GenAI gateway for enabling access to 100s of use cases to internal SLMs, LLMs and commercial LLMs.

4. Aegis : AI Safety layer to moderate incoming and outgoing content for filtering as per customer preferences.

5. Hertz : A model benchmarking platform for certifying functional, safety and non-functional metrics for a model.

6. Maestro : A workflow orchestrator built on top of Airflow for enabling data, machine learning and GenAI pipelines.

7. Gandalf : A complete end to end control plane for accessing GenAI integrations from model foundry, LLM leaderboard and LLM playground.

8. Batch Predict : A batch inferring framework built on top of FDP (Flipkart Data Platform) for offline prediction enabling distributed data processing on Spark cluster.

9. LabelBolt : Data Labeling framework used by Data Scientists and Ops users to label/annotate text, images, audio etc.

10. Ether and Luna(beta) : A framework to build and manage machine learning environments.

11. Model Monitoring (Jugnu) : Capability to continuously monitor models in production with an aim to identify model decay as early as possible.

12. GPU Engineering : Develop frameworks, solutions to utilize GPUs in Flipkart to the maximum extent while model training and model inference.

You are responsible for :

1. Architects at Flipkart are responsible for driving Technology & good practices in Engineering in their respective teams.

2. Actively participate in development along with team members for as much as 75% of their time, creating modules & systems that can then be treated as a working reflection of the best practices.

3. Being responsible for Scaling, Performance & Quality for the team.

4. Setting up best practices to help the team achieve the above and constantly thinking about improving the technology use are your responsibilities.

5. Experiment with new & relevant technologies and tools, and drive adoption while measuring yourself on the impact you can create.

6. Implementation of long-term technology vision for your team.

7. Responsible for the architecture of your platform; ensuring it is aligned to the long term requirements and the platform vision.

8. Active participant in technology forums and represent Flipkart in internal and external forums.

9. Creating architectures & designs for new solutions around existing and new problem spaces at Flipkart.

10. Drive technology & tool choices for your team & be responsible for them.

11. Going beyond your role and contributing to make the organization & business better is an expectation.

You should have :

1. You serve as a technical expert, providing direction and oversight on software development methodologies, technologies, and best practices.

2. You are proficient in programming on either Python or Java.

3. You have experience developing REST APIs, SDKs using Python or Java.

4. You are proficient at developing containerized applications leveraging Docker, Kubernetes or related container stack.

5. You are a champion of SDLC practices.

6. You understand concepts of application development, infrastructure management, data engineering and platform constructs.

7. You are aware of the Machine Learning/Deep Learning development life cycle covering aspects of EDA, Model development, (re) training and inference pipelines.

8. You have experience with Deep Learning tools and libraries such as TensorFlow, Keras, and PyTorch.

9. You have exposure to Data Preparation needs for different use cases. Data preparation involves tasks such as data cleaning, data transformation, feature engineering, and data integration.

10. You have knowledge of storage systems and different Database systems.

11. You are a hustler, tinkerer with open source frameworks.

Educational Qualifications :

1. Bachelors in Computer Science or related field.

2. Must have taken up courses on Machine Learning/Deep Learning/ MLOps or possess a Masters in Data Science or related field.

3. Demonstrable industry or academic experience with Machine Learning/Deep Learning domain. A GitHub repository showcasing portfolio of work on ML/DL Engineering.

Bonus Points if you have :

1. Experience with distributed data processing such as MR, Apache Spark, Apache Beam

2. Understanding of common ML algorithms, Deep Learning, familiarity with NLP or computer vision domain , usage of GPUs in ML lifecycle.

3. Exposure to CUDA programming.

4. Experience with Google Cloud Platform or any other public cloud.

5. Working knowledge of Kubernetes or any container orchestration technology/framework.

What will you do :

1. As a Machine Learning leader, you will play a crucial role in the development and implementation of cutting-edge artificial intelligence products.

2. Lead a team of technical professionals, providing guidance, mentorship, and support to ensure successful project delivery.

3. Your responsibilities will involve designing and constructing platform tools for deploying sophisticated machine learning models, as well as refining and updating existing systems.

4. Build platform solutions to drive greater productivity for Data Scientists at Flipkart.

5. Build tools to support adoption and quick productization of Data Science solutions.

6. Build on recent advancements in the area of Generative AI, LLMs.

7. Build frameworks to support distributed training, large model training within Flipkart on Hybrid compute.

8. Develop services on top of Computer Vision and Natural Language capabilities developed by Flipkart Data Scientists.

9. Develop self serve platform to allow Data Scientists to manage training, retraining, deployment, scheduling, monitoring, and improvement of ML models.

10. Work on problems such as Data Version Controller, Experiment Tracking, Data Labeling, Active Learning, training large language models, building a cluster of GPUs optimized for inferring and training needs, both.

11. Explore and conduct PoC using open source frameworks , AZURE and GCP offerings on Data Science.

Why work with us :

- At Flipkart, we promise to maximize what you care about the most. You'll get an opportunity to leave a mark and create your own legacy, you'll have freedom to experiment, learn and grow, you'll get to work with the best in our inclusive teams and experience our culture of care to ensure that you can focus on doing your best work.

Every Flipster is a Maximiser - Curious, Audacious, Entrepreneurial, Purpose-Driven, Excited By Big Challenges, A Team Player and Inclusive. If this is who you are, then Flipkar is the place where you can thrive.

We understand that your own aspirations and journeys are unique. So you choose what you want to maximize, and we provide you the platform for it. Because when you maximize, we maximize.

Visit our career site to know more about what it means to work at Flipkart.

Flipkart respects and values differences. We are proud to be an equal opportunity employer and are committed to promoting a diverse workplace culture where every individual is valued for who they are, regardless of race, color, age, faith, gender identity or expression, sexual orientation, nationality, marital status, professional pedigree, physical ability, veteran status, or any other characteristics. We firmly believe that when people with diverse identities and perspectives work together, we create the most value - for our customers, our people, and society. We are committed to fostering a culture of inclusion and belonging, where diverse teams innovate and excel every day.