Sep 2021 – Jan 2024 · 2 yrs 4 mos · Bangalore Urban

• ● Built the MLOps platform for aixplain from scratch.
• ● People managed a team for delivering ML Engineering MLOps.
• ● Designing end to end solutions involving ML systems.
• ● Building systems to train and finetune large language models.
• 1. Developed a cost calculator for both training and inference of an LLM.
• 2. A benchmarking report for comparing GPUs and which ones to use for operational costs.
• 3. Building a scalable fully sharded data parallel pipeline for training any LLM with large scale data, checkpointing and memory offloading.
• 4. Inference at scale with a server-less architecture, minimising cold starts to within 30 seconds as server-less architectures introduce cold start times for models of >7B parameters on cloud infrastructure where warm up setup is more challenging.
• 5. Optimised ML models performance with using fast transformer layers and continuous batching techniques on the server apart from standard optimisation techniques like model pruning and quantisation.

Sep 2021 – Sep 2023 · 2 yrs · Bangalore Urban, Karnataka, India

Nov 2020 – Sep 2021 · 10 mos

• Worked as a Data Scientist in Contact Center AI
• Built Agent Assistant System with low latency needs and MRR > 0.3

Jan 2019 – Nov 2020 · 1 yr 10 mos

• Cisco AI Research - CTO team
• Contributor to Kubeflow (www.kubeflow.org)
• Data Science with Natural Language Processing

Mar 2020 – Jun 2020 · 3 mos

• Open Source Privacy based Contact Tracing apps for Healthcare Authorities - initiated by MIT, Boston
• ● Built the first version of the backend for Safe Places tool (https://github.com/Path-Check/safeplaces-backend)

Mar 2019 – Oct 2019 · 7 mos · Bengaluru Area, India

• Helped team Anar from day one on their mission of building a business connection platform:
• ● Technical recruiting
• ● Technical planning and initial deliveries
• ● Advice on tech stack and architecture
• Impact:
• ● This resulted in Anar building a platform hosting over 100,000 businesses today.
• ● Anar has recently announced their seed round of $6.2M with Accel and Elevation partners.

Sep 2017 – Oct 2018 · 1 yr 1 mo

• Platform and Systems Engineering
• ● Engineering solutions for Data Science training workflows
• ● Developed pipelines for Medical Screening Solutions powered by AI

Mar 2017 – Aug 2017 · 5 mos

• Site Reliablity Engineering
• ● Cross-platform Auto-scaling infrastructure with custom metrics

May 2016 – Feb 2017 · 9 mos

• Site Reliablity Engineering
• ● Automated the entire infrastructure requirements from environment deployments, to monitoring
• ● Setup Continuous Integration in our development workflow
• ● Development to deploy time of less than 12 minutes
• ● Automated CI test suite for build quality

Aug 2015 – Apr 2016 · 8 mos · Mumbai

• Data Engineering
• ● Designed and implemented an ETL pipeline for data warehousing and optimized Amazon Redshift as our warehouse, increasing performance by 50%.
• ● Created a Data Lake in Redshift by ensuring replication of data from different sources using Apache Sqoop, which significantly reduced the impact of ETL delivery time with respect to data drift in source schema.
• ● Languages and Libraries used: Apache Hadoop, Pig, Sqoop, Java, Python
• Infrastructure Engineering
• ● Managed AWS Infrastructure for entire tech team and got the cost down by 25% by developing a resource effective deployment pipeline.
• ● Languages and Libraries used: Python, Bash, Ruby, CoffeeScript

Jun 2014 – Aug 2015 · 1 yr 2 mos · Pune Area, India

• Backend Development
• ● Complete owner of, a versioned content API for Vaultize Platform, created in just about 5 months; by extending code serving web templates.
• ● Languages and Libraries used: Python and Web.py
• Android Development
• ● Added several security features to the Android app; i.e. PIN functionality, file encrypt-decrypt and delete based on interaction etc.
• ● Languages and APIs used: Java, Android API 9 to 21

May 2013 – Jul 2013 · 2 mos · Pune Area, India

• Researched about the PST File Format. Developed a generalized B-Tree Module in Python for supporting B- Tree data-structure used by this file format. This module can be used to provide the facilities of extracting mails from and compressing mails to PST files.
• Languages/APIs used: Python