Arjhun Sreedar

I solve problems with code


about me

Hi! I'm Arjhun, a software engineer and researcher currently pursuing my Master's in Computer Science at UMass Amherst. I specialize in suddenly wanting to build something overnight, and losing interest the very next day.

When I'm not busy abandoning yesterday's brilliant idea, I build things that matter. I've professionally worked in climate tech and financial services domains.

Outside of code, I moonlight as a tech writer, open-source contributor and occasional cricketer.

Right now I'm deep in the rabbit hole of LLMs (aren't we all?). Always up to collaborate on interesting stuff! Hit me up at asreedar@umass.edu or arjunsreedar26@gmail.com.

2025 - 2027
Pusuing my Master of Science in Computer Science at University of Massachusetts, Amherst. Hope to do a lot of relevant research!
2024 - 2025
Software Engineer 1 at Oracle. My professional experience at Oracle was quite profound. I learnt a lot about how financial systems were handled a scale!

I significantly improved the debugging experience for our team by integrating CI metrics, like build times, failure trends, and flaky test tracking, directly into our internal portal UI, which sped up issue identification by about 20%.
To further streamline troubleshooting, I built a tool that automatically pulls and aggregates logs from various internal Oracle services, cutting diagnosis time by 30 to 40%.
I also modernized a chunk of legacy React code, which not only reduced page load times by 10% but also led to a noticeable improvement in user experience and made development smoother and more efficient for the team.
2023 - 2024
Founding Software Engineer Intern at CarbonMarketsHQ. Helped the company procure and manage their data, which was foundational to their mission statement. One of my most challending work experiences.

I built and automated a robust data pipeline to handle over 22,000 carbon offset projects and 7 billion credits, pulling from four different data sources.
I also managed the underlying PostgreSQL database and handled CI/CD workflows using Azure.
On the ML side, I developed a multi-agent Retrieval-Augmented Generation (RAG) system powered by Llama 3, optimizing it with LFU caching to improve query speeds by 60%. I extended the system to automatically generate detailed reports and visualizations, and wrote comprehensive load tests that ultimately boosted the overall pipeline performance by 15%.
2022 - 2023
Worked as a Research Intern at Amrita x Sony SSUP funded project focusing on low-cost food drying in a solar polyhouse modelled as a cyber physical system.

Primarily worked on the integration team responsible to making data from the microcontrollers accessible on the cloud.
Redesigned the existing backend module towards a microservices architecture and the entire dashboard frontend.
2020 - 2024
B.Tech in Computer Science and Engineering at Amrita Vishwa Vidyapeetham, Coimbatore. Really fell in love with CS during my time there, but eternally hate the food and campus.
writing
I occasionally write on medium.

→ my content suggestions

projects
jargon is a NER based privacy policy analyser which identifies sections of importance in large jargons of text based on predefined tags generated by LLMs
toolboxAi is a web tool to ask questions about your pdf. It routes between LLMs, maintains conversation history and can be used to ask questions about your pdf, or even specific sections of it.
ChokeSpeare is a collection of different generative models trained on different kinds of poems and can generate poems in a mixed style.
iDEA community website is a pure react website for the undergrad club I co-founded. It does not use any external libraries except styled components. An unorthodox design and code structure. Find the website hosted here.
Reddit Scraper is a CLI tool that can scrape reddit content quickly using a unique checkpointing mechanism and JSON trick.
WeisfeilerLehman Quick Compute is a python script that can speed up WL Kernel computations by computing as batches and merging.
misc
OSNEM 2025 (In peer-review)
Arjhun Sreedar, Pranav Deepak, Dayanand V
Medium
600+ claps and 30k+ reads
Leetcode
250+ problems solved
Finalist
CoinSwitch Kuber Blockchain Hackathon 2022
Finalist
Hack@Arch Hackathon 2022
MxFlow
Active contributor at MxFlow, an AI tool for email based workflow automation.
Pandas
Active contributor in the pandas repo, including triaging and bug fixing.
contact

Have something interesting to discuss? Drop me a message.