Roshan Kumar

Proof of Experience

Goldman Sachs · Jan 2025 – Present

Built a logic-driven financial agent utilizing LLM context injection and financial entities recognition within user queries, dynamically pulling expert research frameworks to provide subject matter expert (SME) level reasoning.
Achieved a 10x QoQ increase in CTR by building an LLM-driven RAG and Entity Recognition service, surfacing highly relevant market articles and research.
Expanded the digital storefront by designing and launching a new ETF product page, increasing the revenue channels and increasing visibility for institutional buyers.
Built a comprehensive real-time reconciliation platform integrating with the balance sheet to ensure data accuracy and regulatory compliance across various asset classes.
Technologies: Python3, Java, TypeScript, FastAPI, Conduit, VectorDB, Grafana, Prometheus, LangChain, GenAI, RAG, Agentic AI, DynamoDB, React, MobX, Vert.x

Goldman Sachs · Jun 2022 – Dec 2024

Automated 5+ Institutional Reports saving 3-7 days of human work on each of them.
Designed and implemented dynamic infrastructure for automated interest rate risk calculation across all financial books, decreasing resource use by 80% and enabling real-time risk assessment for trading desks.
This also decreased the overall reporting cycle from 65 days to 7 days.
Enhanced the PnL dashboard by implementing view saving and browser-based caching, allowing users to save and share frequently used configurations.
Technologies: Python3, Java, FastAPI, Conduit, Grafana, Prometheus, DynamoDB, Plotly

Tekion Corp. · Jan 2022 – May 2022

Developed scalable backend architecture for work management tool with seamless Jira integration using webhooks for real-time ticket synchronization across multiple development teams
Technologies: Spring Boot, MongoDB, Elastic Search

No completed sessions yet. Anything Roshan finishes shows up here with the evidence behind it.

Indian Institute of Information Technology, Guwahati · 2018 — 2022