Résumé

• Designed and implemented a natural language–controlled browser agent capable of autonomously performing multi-step web tasks.
• Led the migration of the Actions Hub to a modular architecture, decoupling action registration from agent execution.
• Integrated Langfuse to enhance observability across LLM systems, implemented RBAC for access control and compliance, and built automated workflows to streamline onboarding of new organizations.

• Led SDK development for PhonePe's B2B payment gateway. Designed and implemented the Java SDK, wrote the CI/CD pipelines, and mentored development of SDKs for other languages. The SDK family has carried tens of crores in total payment volume.
• Owned the Epoch framework for online merchant services. Introduced structural reforms for its scripts and fixed a long-standing org-wide missing-logs issue. Wrote reconciliation and reporting scripts to analyze and close pending transactions.
• Integrated disaster recovery changes into online merchant services and led the DR drills end to end, coordinating with SREs and other stakeholders. Upgraded infrastructure and migrated the services to the Mahape data centre. (PhonePe's engineering blog describes what a drill of this kind involves at their scale).
• Designed and implemented end-to-end accounting and template management for partner integrations. Added backend error codes, events for PG offerings, and merchant-level kill switches on live traffic.

• Associate in the Fixed Income Division's Macro Valuation Adjustment team, improving front-office risk systems.
• Led implementation of 10+ valuation adjustments across desks (structured notes, FX), automating accurate VA reporting to streamline controls, reduce cognitive load, and save 25 engineer-hours per month for the Valuation Control and Strats teams.

• Implemented a cost-efficient heating solution using machine learning for Energy.AI clients. Conducted preliminary research and prototyped an agile methodology incorporating genetically evolving algorithms, with an anticipated 40–50% decrease in client costs.

• Thesis placement from BITS Pilani. Wrote Python scripts simulating user behaviour and created the reinforcement learning environment for research project PAL, which aimed to provide personalized learning for its users. Created augmented data for training algorithms.

• Thesis placement from BITS Pilani. Also worked as a research assistant for the UNSW Business School alongside it, developing a system to assess event impacts on assets under Professor Fethi Rabhi's guidance. Contributed to research on dark pool and hedge fund activity at the ASX, analyzing trading patterns and fund distribution across Australian exchanges. Part of this work is open source: RA_Bschool.

• Created a one-stop solution for stakeholders to view strategy performance, writing Python scripts to analyse existing strategies. Developed algorithms for Indian equities with a Sharpe above 4 before trading costs.

• Developed alphas based on technical and fundamental analysis as trading signals. Selected alphas were traded in the market.

• PhonePe/phonepe-pg-sdk-java — Author. The public Java SDK for PhonePe's B2B payment gateway: the developer-facing API, the release workflow, and the CI that ships it. 40 commits
• PhonePe/epoch-cli — Author. The CLI for Epoch, the cron for the Drove container orchestrator: cluster management commands, pause-all, topology filtering, and removing the PhonePe-specific constructs so the tool could stand on its own. 27 commits
• pyjanitor — Contributor. Clean APIs for data cleaning in pandas, 1.5k stars. Two merged features: drop_constant_columns and a reset-index flag on row_to_names. 2 merged PRs
• RA_Bschool — Author. Research assistant work for the UNSW Business School, 2019.

Coursework: data structures and algorithms, object-oriented programming, database systems, operating systems, computer architecture, machine learning, neural networks and fuzzy logic.

• Terminal One — MIT vs Harvard (2020) — Runner-up. Run by Citadel and Correlation One, across a field of 30,000+ engineers and data scientists. Represented MIT against Harvard as a team of two with Akash Vaish. Built Go Get It (go1, go2, go3 by the end), an algorithm that plays Citadel's Terminal against other algorithms, adapting its offence to the opponent's attack. Reached the final and lost 5–4. Write-up · certificate.
• Auquan Spring Challenge (2019) — Winner. Scored above 0.03, placing in the top prize bracket, with a cash prize. Logistic regression with hyperparameter tuning performed best. Write-up · confirmation.
• WorldQuant Spring Alphathon (2017) — Gold, Level 2. Ranked 23rd internationally and 2nd in India, leaderboard score 32,091. Led directly to the research consultant offer at WorldQuant. Write-up · certificate · leaderboard.