Experience
2024 — Now
2024 — Now
London, England, United Kingdom
Deductive AI is an applied AI research lab building AGI to enable self-healing software systems.
Our code-aware observability platform helps numerous companies root-cause and mitigate large-scale software outages by reasoning about distributed systems, code, and statistical anomalies in real time across unprecedented volumes of structured and unstructured data. We’re a team of engineers and researchers with decades of experience building and maintaining large-scale production systems at Databricks, Facebook, ThoughtSpot, Google, Splunk, and Amazon.
2019 — 2024
2019 — 2024
London, United Kingdom
Data Infrastructure (Presto, Spark):
• Presto and Velox contributor https://github.com/prestodb/presto, https://github.com/facebookincubator/velox
• Working on enabling Presto run warehouse sized big queries using Velox, a C++ vectorized database acceleration.
Ads Infra:
• Led a team of 8 engineers to achieve efficiency improvements of > 2x resulting in a saving of ∼$40M in Opex cost.
• Modernised and revamped Ads Experimentation data platform.
2016 — 2019
2016 — 2019
Singapore
* Forked and enhanced a golang based circuit-breaker library with features to improve resilience, in some cases it has reduced up-to 40% of the errors. It is one of the most widely used library within Grab, and is open sourced at https://github.com/myteksi/hystrix-go
* Designed, developed an in-house configuration management system. The configuration system was designed specifically to solve for strong type-casting, validation, and was hierarchical to support multiple environments/clusters. The system also worked as a distributed log service for all changes going to production.
* Working on building a multi-tenant Segmentation Platform primarily for the use of business teams and data-scientists. It will help them create user segments based on various parameters and models in batch and streaming modes.
2016 — 2018
2016 — 2018
Amsterdam Area, Netherlands
Developing SaaS solution for neural art for images and videos.
It uses deep learning technique to learn from an existing style image and applies the same onto another. It uses distributed systems and multi-GPU architecture to improve computation time.
I also do sales and marketing for the product.
NeuralPainter twitter bot - https://twitter.com/neuralpainter
2015 — 2016
2015 — 2016
Worked with partner security team. During my stint at Booking.com, I developed various tools to help improve the security of the system like email campaign detection and spam detection system.
Education
Birla Institute of Technology, Mesra