Jun 2019 – Jan 2021 · 1 yr 7 mos · Greater Bengaluru Area · On-site

• I joined Confluent pre-IPO, contributed to its growth as it went public, working on distributed data infrastructure.
• ● Responsible for designing and implementing a set of Smart Notification APIs for Confluent Cloud to proactively alert customers of potential issues with their Kafka deployment.
• ● Developed a framework to guarantee SLAs for the org-wide logging pipeline, ensuring completeness & robustness.
• ● Led the unit & performance testing, product documentation, and official release of the Redis sink connector for Kafka.

Oct 2017 – Present · 8 yrs 5 mos · Remote

• ● Invested in 15+ startups across multiple funding stages, from seed to growth. Typical check sizes range from $5K–$10K for early-stage opportunities to $50K for mid-stage ventures. Focused on scalable business models, strong founding teams, and data-driven market opportunities.
• ● Mostly interested in the domains of blockchain, AI, and security.
• ● Notable investments: Tron, Wanchain, Evenflow, Agnikul Cosmos.

Oct 2017 – Aug 2025 · 7 yrs 10 mos · India

• ● I’m a self-taught trader. After winding down my first startup (MadX) in 2017, I began exploring trading, a journey that started with big wins and bigger lessons.
• ● From 2019 to 2021, I worked at Confluent (Software engineering) before returning to trading full-time in early 2021, this time with a disciplined and systematic approach.
• ● From 2021 to 2025, I developed and refined a profitable trading bot that integrated fundamental and technical analysis to execute custom trend, momentum, and swing strategies with controlled risk. The system currently runs live across multiple liquid futures and spot markets.

Nov 2016 – Sep 2017 · 10 mos · Greater Delhi Area

• ● MadX Ads is an Internet solutions company focused on developing innovative and high-impact solutions for premium digital publishers and advertisers.
• ● We believe in providing value to our clients in the form of targeted, contextually relevant, unobtrusive, cognitively engaging and high customized advertisement solutions across search feeds, in-image, in-video and rich media formats.
• ● Our proprietary machine learning technology and our advanced set of user tracking methodologies, combined with our attention-commanding novel ad formats relate to the end users on a deep personal level, maximizing returns for all parties involved.

Jul 2014 – Oct 2016 · 2 yrs 3 mos

• ● Developed from scratch, a framework for bid price optimization in real-time bid auction mechanisms in contextual advertising using reinforcement exploration-exploitation optimization methods.
• ● Led a team of 4 colleagues to design a smart data analytics framework that provided non-trivial insights about data with minimum user input, including cluster analysis and anomaly detection, custom tailored for, but not limited to, attributes and measures present in typical ad-exchange networks.
• ● Problem Setter / Technical Interviewer for Recruitment Team: Interviewed 200+ candidates for positions in Software Engineering, Software Development and R&D (freshers as well as lateral).

Jun 2013 – Jun 2014 · 1 yr

• ● Identified bottlenecks in the real-time bidder transaction path for advertisement networks by micro-benchmarking and profiling. Designed a stratified internal architecture to improve performance, scalability and extensibility.
• ● Resolved undiscovered vulnerabilities in the system by simplifying the underlying logic and eliminating unnecessary levels of complexity. The strategies led to a 50% increase in the throughput (serving over 100k+ qps) of the system.

May 2012 – Jul 2012 · 2 mos · Lausanne Metropolitan Area

• "Refinement of the VPR packer for the AIC-based FPGA"
• ● Worked on a novel alternate architecture for the FPGAs (Field-Programmable Gate Arrays), called the AIC’s (And Inverter Cones) which are based on the AIG’s (And-Inverter Graphs) and have a complexity that is linear in the number of inputs.
• ● Modified the VPR design flow by improving the critical path delay. The new design flow reduced the order of runtime for the packing of the benchmark blif file from days to minutes.