Anish Kelkar
Software Engineer
Bengaluru
Summary
Curious and passionate software developer with 6+ years of experience as a full-stack software engineer/ game developer/ machine learning compute engineer. Skilled in multiple programming languages, software design, and debugging. Strong focus on team collaboration and achieving results. Reliable, adaptable, and solution-focused.
Overview
8
8
years of professional experience
Work History
Software Engineer 2 (Machine Learning Backend Engineer)
PayPal
12.2022 - Current
- Worked on developing model inference engine where all Risk Models are computed. Also developed pipeline for automated model deployment in shadow production environment for pilot testing.
- Reduced risk model computation latency by 20% by redesigning a distributed compute service (20k variables/request, 600 ML models).
- Designing distributed in-memory database layer (called Juno) to reduce Database load and using logs analysis with PySpark to improve Cache Warmup to identify keys.
- Performing Distributed System debugging and improving availability by using understanding bottlenecks of JVM run-time and using ZGC over G1GC
- Optimized memory usage by 15% through profiling legacy system patterns (VisualVM, JMeter), resulting in $15K/year infrastructure savings.
- Optimized the Java build process to cut service startup time by 50%, minimizing downtime during twice-weekly releases across 1000 servers
- Using GCP-BigQuery to store BrainTree dispute information and merchant contracts for quick loss-cap validation and dispute resolution.
- Use React website to develop charge-back protection products for BrainTree Merchants
Software Engineer (Full Stack Game Developer)
Zynga
05.2021 - 08.2022
- Developed Slot Machine game and developed features like Daily Challenges and implemented Favourite functionality
- Used client-side gameplay in C# (Unity) and built backend services in Java on AWS, ensuring smooth integration between client and server
- Remote, India
Software Engineer (Backend Engineer)
JP Morgan Chase
07.2019 - 05.2021
- Modernized a customer-facing, XML-based monolithic transaction service by migrating to microservices compliant with FDX standards
- Handling of 50,000+ daily requests for requesting details for transaction and account balances. The data is sent to 3rd party aggregators.
- Bengaluru -> Remote, India
Software Engineer Intern (Data Analyst and Android Engineer)
JP Morgan Chase
05.2018 - 07.2018
- Analyzed Chase Mobile app crash logs using Python (pandas, scikit-learn) to identify failure patterns; insights from this NLP analysis guided report generation for the team.
- Bengaluru, India
Education
Bachelor of Technology - Computer Science and Engineering
Visvesvaraya National Institute of Technology
Nagpur, MH05.2019
Skills
Programming Languages: Rust, Python, Java, C, C, C#, Bash
Affiliations
II. Opensource Contribution to Spring framework (reactor-pool)
- HTTPS://GITHUB.COM/REACTOR/REACTOR-POOL
- worked on this open source contribution from June 2024 to July 2024
- This is project is part of the Apache-2.0 Licence
- This Library focuses on managing reusable resources (e.g., database connections, network sockets) in a reactive, non-blocking manner.
- My contribution improved maintainability by adding coding standards to AbstractPool component.
Timeline
Software Engineer 2 (Machine Learning Backend Engineer)
PayPal
12.2022 - Current
Software Engineer (Full Stack Game Developer)
Zynga
05.2021 - 08.2022
Software Engineer (Backend Engineer)
JP Morgan Chase
07.2019 - 05.2021
Software Engineer Intern (Data Analyst and Android Engineer)
JP Morgan Chase
05.2018 - 07.2018
Bachelor of Technology - Computer Science and Engineering
Visvesvaraya National Institute of Technology
Research Publications
- DOI: 10.32908/ahswn.v55.5829
- JournalName: Ad Hoc & Sensor Wireless Networks
- Developed distributed algorithm for tiger census counting under strict memory constraints (5MB per node) using C and Python, to aid tiger census planning.
- C99 was used to write the communication algorithm between the simulated nodes.
- Python was used to plot the performance to measure between two parent consensus algorithm.
- Use of Bash scripting, to filter on logs written to be used by python. to remove redundant actions mentioned in the logs.