About Me
Building Digital Experiences
I'm Sushvin, a Full Stack Developer and AI Engineer from Hyderabad, focused on building fast, responsive, and user-friendly applications across web development and applied AI. I love turning ideas into scalable, intelligent digital experiences.
AI/NLP & Full-Stack Engineer
Sushvin Marimuthu
About Me
My journey as a full-stack developer began in 2022 with freelancing for clients across India, where I built end-to-end solutions and collaborated closely with clients. I later joined Mallow as a Project Intern (Backend Developer), which helped sharpen my backend skills and introduced me to structured, team-based development. After that, I interned at IIIT Hyderabad and subsequently became a Junior Software Engineer at LTRC, contributing to full-stack tools for linguistic research. A year later, I received an offer to pursue a Master’s by Research, which I’m currently in my second year of. Alongside academics, I continue to work on app development and AI research, blending innovation with practical impact.
My Journey
Bachelor of Engineering in CSE @ Karpagam institute of Technology
- Oct 2021 - Feb 2023 Project
Karpagam e-Governance Portal - Centralized Academic Management System- Developed an academic management system for student records, attendance, schedules, and grading, with role-based access control and background jobs (Redis Queue) for asynchronous tasks.
- Implemented leave modules for students and faculty with online applications and approvals, synced with attendance such that approved leaves are reflected automatically.
- Mar 2022 - Aug 2022 Project
Smart Suit: Developed an IoT device and web application to enhance coal miner safety for Smart India Hackathon 2022.
Project Intern @ Mallow Technologies
- Feb 2023 - May 2023 Project
Traced and fixed a backend logic bug in the GIST CRM chat module that caused incorrect customer and message counts in production, restoring data accuracy across chat workflows.
Junior Software Engineer @ LTRC, IIIT - Hyderabad
- Sep 2023 - Jul 2024 Project
Web Development: Designed and developed websites for LTRC events, including IASNLP and PLO, as well as course pages, supporting event outreach and academic learning. - Sep 2023 - Feb 2024 Project
Tamil Syntactic Parser: Fine-tuned the Trankit framework on an in-house annotated Tamil data to build an AI-powered Tamil syntactic parser that decomposes complex sentences and maps word relationships for Tamil language research which is presented at the Kani Tamil 2024 conference. - Dec 17, 2023 Publication
Disfluency Identification for Indian Languages using Transformer-based Models: Trained a model to detect and classify disfluencies in text, presented at ICON 2023 - Jun 21, 2024 Publication
Medical Visual Question Answering with Vision-Language Models: Our work for the MEDIQA-M3G 2024 shared task uses a fine-tuned ViLT model for clinical dermatology VQA, ranking 6th on the leaderboard. Post-challenge, we trained on more data and explored large VLMs such as Gemini and LLaVA.
MS by Research Student @ IIIT - Hyderabad
- Aug 2024 - Present Research
Researcher @ LTRC, IIIT - Hyderabad
Enrolled in MS by Research at IIIT-Hyderabad, working on Medical NLP research and building practical applications in the field. - Aug 2024 - Dec 2025 Project
MediHelp AI (for CMC Vellore Doctors)- Built a real-time collaborative, multi-format document editor (Tiptap, WebSockets) and moved heavy operations to asynchronous queues (BullMQ) for better performance and reliability under load, reducing manual document work by ~60%.
- Integrated LLM-powered translation, Q&A, summarization, and text-to-table features to support doctors' report writing.
- Deployed the application using Docker and set up a CI/CD pipeline with GitHub Actions that deploys to AWS (EC2) for staging and CMC's internal infrastructure for production.
- May 4, 2025 Publication
SpanSense - Perspective-specific span identification and Summarization
Second Workshop on Patient-Oriented Language Processing (CL4Health), NAACL 2025- Secured 6th place on the official PerAnsSumm 2025 leaderboard with an overall score of 49% for perspective-aware medical CQA summarization.
- Designed a RoBERTa-large + BART-large pipeline for perspective span classification and summarization, achieving 90% Macro-F1, 92% Weighted-F1, and 84% BERTScore.
- May 2025 - Sep 2026 Project
Structured Radiology Intelligence (SRI):- Improved clinical data standardization for rectal cancer MRI reports by fine-tuning LLaMA 3.1 8B Instruct for structured information extraction and building a curated domain-specific dataset using multiple LLMs.
- Developed a web application for dataset annotation, validation, quality control, and demo, and deployed the fine-tuned LLM as a scalable inference service with guardrails for reliable structured MRI report processing.
- May 2026 Publication
Structured Radiology Intelligence: Extracting structured data from MRI Reports using LLMs
Third Workshop on Patient-Oriented Language Processing (CL4Health), LREC 2026- Achieved an 87.36% composite score and 18.42% performance improvement using LLaMA 3.1 8B Instruct for automated MRI report structuring.
- Developed and benchmarked a 55-variable clinical dataset, evaluating LLaMA, Qwen, and DeepSeek through prompt-tuned and fine-tuned LLM pipelines.
Full-Stack Developer @ Rubrix.ai
- May 2026 - Present Research Project Internship
- Developed an AI-powered semantic recommendation system that matches online resources to course content and created a database of resources with precomputed similarity scores, surfacing matches at ≥ 0.85 similarity. It replaced a paid Perplexity API workflow, saving ~₹10K/month.
- Implemented code and text plagiarism detection for student submissions, triggered automatically by a Celery worker after every submission, replacing manual integrity checks.
- Built an automated question paper generation feature that lets instructors define templates such as sections, marks per question, units covered, number of sets, and AI-generated vs. previous-year question ratios, eliminating manual assembly and reducing API costs by ~50%.
Projects
SRI
May 2025 – Sep 2026
Structured Radiology Intelligence (SRI) is a schema-guided system that converts unstructured and semi-structured rectal cancer MRI reports into structured data.
MediHelp AI
Aug 2024 - Dec 2025
A document editor that extracts text from .docx, .pdf, .txt, and .rtf files, allowing users to edit, translate, and perform Q&A tasks.
Rasa
Jun 2025
Rasa is a Sanskrit annotation tool for labeling utterances and context, enabling accurate linguistic and semantic analysis.
SRG-Journal
May 2025 - Present
SRG-Journal is a tool for chemistry researchers to manage project and experiment details, along with additional features.
Tamil Parser
Sep 2023 - Feb 2024
A parser is an automated tool that provides the internal structure of sentences. It decomposes complex structures into their constituent parts and represents relations between words.
Publications
My research publications
2026
Structured Radiology Intelligence: Extracting structured data from MRI Reports using LLMs
Sushvin Marimuthu and Parameswari Krishnamurthy
10.63317/4vob5zztrfso2025
SpanSense - Perspective-specific span identification and Summarization
Sushvin Marimuthu and Parameswari Krishnamurthy
10.18653/v1/2025.cl4health-1.372024
Medical Visual Question Answering with Vision-Language Models
Jerrin Thomas and Sushvin Marimuthu and Parameswari Krishnamurthy
10.18653/v1/2024.clinicalnlp-1.67Connect with me
Get in touch to discuss your next project
Are you ready to take your project to the next level? Whether you’re looking for a new website, a web application, or simply need advice.