Aug 2021 – Present · 4 yrs 7 mos · Bengaluru, Karnataka, India · On-site

• Bootstrapped GTF data science platform team, scaled the team to 6 people. Dual hatted the role of ML engineer and MLOps engineer to support data science teams of GTF.
• As an MLOps engineer doing following:
• Setup and maintaining Kubeflow Platform on GCP with multi node disk support via NFS server. Setup pipelines for deployments of NFS servers, Kubeflow custom notebook and pipeline images, setup convention for DSP for accessing the kubeflow features. Added support of Elyra for faster development of DS pipelines for training and batch predictions via KF pipelines.
• Maintaining Merlin/CaraML platform for deploying Data Science Models faster.
• Building inhouse Self Serve Feature Engineering solution for realtime models serving, training and monitoring. This has increased model deployment output drastically
• As ML Engineer
• Have deployed more than 5 different realtime models in production having more than thousand features with different category of features ranging from tabular features to graph features in subsecond latencies.Came up with innovative solutions to leverage the model results in realtime.
• Setup monitoring for concept and data drift for realtime model.

Sep 2019 – Jul 2021 · 1 yr 10 mos · Bengaluru, Karnataka, India · On-site

• Worked in GTF Sanctions Platform, Helped in setting up tooling for running rules to prevent fraud, taking sanctions in real/near realtime. Majorly contributed in following systems:
• User Based Authorization Server: Developed UBAC service which implements Envoy's ext-autz filter to block apis calls using api routes and header values. This was the first layer to take sanction against user.
• Fraud Rules System: Developed FRS which helps in running different types of rules in real/near-realtime to prevent fraud which involves longer time period using Redis and Clojure.
• Sanctions Service: Developed Sanctions service, which helped in taking temporary and permanent sanctions, along with repeal flow. Used quartz, clojure and kafka to achieve this.

Aug 2018 – Jul 2019 · 11 mos · On-site

• Worked with GTF convenience team, which took care of increasing convenience of using of GoPay wallet. Contributed in Direct Debit feature of gopay app, added support of payment switch in gofood flow for gofin vs gopay payment options. As part of this, also deprecated existing legacy system apis related to payment switch.
• Also contributed consul-envoy-xds i.e. XDS api control plane for consul for bidirectional updates of endpoints(EDS). Built internal tooling to generate gRPC clients with internal libraries which included NR segments, logging, connection pooling, service discovery features.

Jul 2017 – Jul 2018 · 1 yr · On-site

• - Worked with GoPay liquidity team which dealt with increasing liquidity of gopay wallet. Responsibility included maintaining bank integration for VA topups, adding support of driver topup via customer wallet etc.