Hey, I'm Dyuti. I'm a Software Engineer with a Master's in Computer Science from Arizona State University, and I just wrapped up my time there as a Graduate Service Assistant, where I got to mentor students through C++, assembly, and embedded systems. Before ASU, I did my Bachelor's in Computer Science and Engineering at Gokaraju Rangaraju Institute of Engineering and Technology, graduating with distinction. Along the way I've built a lot of things I'm proud of: serverless pipelines on AWS, scrapers that pull structured data out of messy web pages, and full stack apps with the MERN stack. As a Software Engineer Intern at Dark Alpha Capital LLC, I built internal tools, data pipelines, and GUIs that people on the team actually used day to day. At Techmirus Innovations, I worked as a Full Stack Development Intern, building responsive, secure apps as part of an Agile team. I work mostly in Python, C++, JavaScript, React, Next.js, Node.js, and AWS, and I genuinely enjoy the process of turning a messy problem into something clean and working. Right now I'm on the job hunt, looking for a team where I can dig into interesting problems and keep growing as an engineer. When I'm not coding, I'm usually traveling somewhere new or making art, so feel free to poke around the rest of the site.
C++
Python
JavaScript
TypeScript
HTML
CSS
SQL
Java
ReactJS
Next.js
Node.js
Kafka
JUnit
Redux
MySQL
MongoDB
PostgreSQL
AWS
GCP
Git
GitHub
Docker
Selenium
Postman
Linux
JIRA
Figma
Master of Science
Computer Science
Arizona State University
Bachelor of Technology
Computer Science and Engineering
Gokaraju Rangaraju Institue of Engineering and Technology
High School
FIITJEE
Back-End Developer
- Built the backend for Mystic Die Studios' internal Employee Dashboard from the ground up using Django, Django REST Framework, and MySQL
- Replaced what used to be manual, scattered tracking with one central system managers and staff can actually rely on
- Designed and secured REST APIs with JWT authentication and GitHub OAuth for user sign-in
- Implemented role-based access control so every user only sees and touches the data their role is supposed to
- Working side by side with frontend engineers to ship scalable portal features using React, Vite, Tailwind CSS, and Docker
- Going back through the backend architecture regularly to tighten up performance and keep the codebase easy to extend as new features get added
Research Aide (Volunteer) – AI Tools & Automation
- Building an agentic AI grading pipeline on LangChain and GPT-4o for automated assignment evaluation
- Built dynamic rubric parsing so the pipeline adapts to whatever grading criteria an instructor sets, instead of relying on one fixed template
- Added multi-format document ingestion, handling PDF and DOCX submissions across a range of subjects
- Designing an LLM-powered personal assistant built on Claude Sonnet with Google Calendar integration
- Handling automated scheduling, task prioritization, and deadline tracking for grading, taking a chunk of the administrative overhead off instructors' plates
Teaching Assistant – Programming & Computer Architecture
- Built C++ autograders from scratch for 8 course sections across 3 semesters
- Took a grading process that used to take instructors 18+ hours per cycle and got it down to under 30 minutes
- Freed up a significant amount of instructor time that could go back into actual teaching
- Debugged over 1,000 low-level issues in student C++, Assembly, and embedded systems code, including memory corruption, undefined behavior, and race conditions
- Used the patterns found while debugging to make the grading pipeline more stable under heavy load from hundreds of submissions at once
Software Engineer Intern
- Designed and shipped a multi-tenant ticketing platform in under 3 weeks using Next.js, PostgreSQL, Docker, and the Gemini API
- Delivered AI-generated ticket summaries along with a shared space for the team to collaborate, for a fintech hedge fund client
- Built 25+ web scrapers with Selenium and BeautifulSoup to automate data collection across complex sites
- Designed fault-tolerant data ingestion pipelines with retry and backoff logic plus rate limiting
- Turned over 600 company pages into clean, analysis-ready PostgreSQL datasets that the research team could query and trust
Full Stack Developer
- Built a MERN-stack employee tracking platform used by 100+ people day to day
- Tuned MongoDB schemas and rethought API calls so the app held up under heavy concurrent use instead of slowing down as more people logged in
- Kept CI/CD pipelines and Git workflows running smoothly as part of a 6-engineer Agile team
- Worked through the full deployment lifecycle together with the team, helping ship more reliably with fewer last minute surprises
Designed and implemented a scalable, serverless face recognition pipeline using AWS Lambda, Docker, and ECR. Deployed containerized Lambda functions for face detection and recognition with PyTorch models, integrated SQS for asynchronous processing, and optimized Docker images for low latency. Achieved 100% success rate with <2s average latency on 100 inference requests.
Developed a full-stack, scalable face recognition web app using AWS IaaS services. Implemented a Python-based HTTP server on EC2 for image classification, S3 for storage, SQS for queuing, and SimpleDB for lookups. Integrated PyTorch models for real-time recognition and deployed an autoscaler managing 15 EC2 instances with <1.0s response latency for 100 concurrent requests.
A full-stack travel application inspired by Airbnb that lets users discover, create, edit, and review travel listings. Features include authentication, reviews, Cloudinary-powered image uploads, and secure CRUD operations for listings.
Independently designed and developed a full-stack, responsive website for improved online presence and patient engagement. Built the frontend with React and Tailwind CSS, ensuring seamless navigation and cross-browser support. Developed RESTful APIs with Node.js and Express.js to handle secure patient appointments and inquiries.
Performed sentiment classification on Amazon reviews using a Random Multimodel Deep Learning (RMDL) architecture with DNN, CNN, and RNN models. Incorporated BERT tokenization and aspect term extraction for preprocessing. Developed a novel Coot-Political Algorithm (CPA), a hybrid optimization technique combining COOT and Political Optimizer, to enhance weight optimization.
Implemented a CNN-based system to improve accuracy in detecting offline signature forgeries. Extracted features such as ratio, centroid, eccentricity, and skewness, and developed a Tkinter-based GUI application to visualize preprocessing stages and final classification. The system successfully identifies genuine vs forged signatures.
Contact
I'd love to hear from you!
Whether it’s an idea, opportunity, or just a friendly hello, I’d be glad to hear from you.