Hi, I'm Kashish
Computer Vision & Machine Learning Developer looking for Summer 2026 internship opportunities
Introduction
Overview
I'm a Computer Vision graduate student at Carnegie Mellon University (Robotics Institute, SCS), passionate about AR/VR, 3D Vision, Vision-Language Models (VLMs), multimodal understanding, and generative visual intelligence. My journey spans both research and industry—shaping ideas into impactful systems.
Previously, I've worked as a research intern at CMU and VJTI, and as a Software Developer Intern at Myntra. My projects explore diverse CV problems—from fine-tuning Stable Diffusion for zero-shot 3D segmentation in cryo-ET, to modeling cell morphology via unsupervised learning, to scaling virtual try-on systems with garment segmentation and LLM-powered recommendation. My work has been published in Springer LNNS and bioRxiv.
Beyond research, I've built production-ready systems: phishing detection pipelines as Chrome extensions, LLM-augmented search at Myntra, and OCR for low-resource scripts using Kolmogorov–Arnold Networks. I enjoy blending classical CV, deep learning, and prompting to solve real-world challenges.
I'm particularly excited about advancing semantically aligned perception, OCR for underrepresented scripts, and applied VLMs in image-based search, AR-powered experiences, and privacy-conscious, real-time vision systems—especially in e-commerce, accessibility, and consumer tech.
What I have done so far
Work Experience.
Computer Vision Research Intern
Carnegie Mellon University
- Delivered near human-level zero-shot segmentation on 3D cryo-ET by fine-tuning Stable Diffusion (LoRA) and applying multi-scale spectral clustering, outperforming unsupervised baselines without annotations.
- Segmented nuclei in gigapixel RNAscope images (IoU 0.84) with fine-tuned CellPose; modeled morphology via unsupervised deep learning and linear mixed models to reveal neuron-type and age-related patterns.
- Working under the guidance of Dr. Min Xu in the Xu Lab, focusing on cutting-edge computer vision applications in computational biology and cryo-electron tomography.
Software Developer Intern
Myntra Designs Private Limited
- Led auto-suggest integration for Luxe store, building a scalable config-driven framework replacing hardcoded logic; drove 23.7% adoption, 46% search growth, and cut future integration effort by 75% through cross-functional collaboration.
- Initiated and drove improvements in typo correction in auto-suggest via advanced heuristics and BM25, reducing latency by 40% and boosting suggestion accuracy.
- Pruned 35% of noisy auto-suggest queries by LLM-based few-shot prompting (96% accuracy) in collaboration with the Product Manager to fix incomplete suggestions that hurt product discovery, boosting search relevance and engagement.
Computer Vision Research Intern
Veermata Jijabai Technological Institute
- Built a real-time Chrome extension for phishing detection, combining visual, HTML, and URL cues; leveraged multi-scale feature extraction (SIFT, ORB, LBP, wavelet hash) and Mask R-CNN for semantic segmentation.
- Integrated an ensemble voting classifier achieving 95.4% accuracy; communicated results through a peer-reviewed manuscript accepted at International Electronics Symposium 2025.
- Collaborated with Dr. Faruk Kazi in the CoE-CNDS lab, developing multimodal phishing detection systems using classical computer vision and deep learning approaches.
Teaching Assistant
Dwarkadas J Sanghvi College of Engineering
- Served as a Teaching Assistant for Prof. Aniket Kore, contributing to the effective delivery of course content.
- Supported students in their learning journey by providing guidance and clarification on course material.
- Conducted Python laboratory sessions for Second Year Computer Engineering students.
- Prepared students for practical exams and final projects by assisting them with hands-on programming exercises and troubleshooting.
Software And Machine Learning Developer
Nextgen Techno Ventures Private Ltd
- Developed an Automatic Face Recognition System using OpenCV, enhancing face detection accuracy and attendance tracking.
- Built an OCR-based Social Media Analysis tool utilizing Pytesseract for text extraction and analysis.
- Created a Car Number Plate Recognition system for compliance checks using TensorFlow and YOLO v8, improving automated vehicle monitoring.
- Developed ShopGPT, a Virtual Shopping Assistant Chatbot leveraging NLP and web scraping tools, and also created a Python library for trading strategies.
My work
Academic Papers.
This section showcases my academic contributions, highlighting my research efforts and findings. Each paper has either been published/accepted or under review, links to their respective PDFs are provided, illustrating my ability to conduct in-depth research and present findings effectively.
A Multimodal Framework for Deepfake Detection
Published in Journal of Electrical Systems 20-10 s (2024): 4772-4787, SCOPUS Indexed
#VGG19
#Mel-Spectograms
#ANN
Detecting Polycystic Ovary Syndrome Through Blending Ensemble Method
To be published in Lecture Notes in Networks and Systems - Springer Book series, index by SCOPUS. Accepted and presented in 6th International Conference on Data & Information Sciences(ICDIS-2024)
#BlendingEnsemble
#ML
#DL
Predicting Solar Energy Generation with Machine Learning based on AQI and Weather Features
Accepted and presented in AISD 2024. Awarded 2nd Best Paper at the workshop.
#LSTM
#ZeroInflated
#PowerTransform
Infectious Disease Forecasting in India Using LLM’s and Deep Learning
Presented in 3rd International conference on Machine Learning and Data Engineering (ICMLDE 2024). Under Review in Procedia Computer Science Journal
#LLM
#NLP
#DL
Cyberbullying or just Sarcasm? Unmasking Coordinated Networks on Reddit
Published in International Journal of Engineering Research in Computer Science and Engineering.
#BERT
#RoBERTA
#GraphNetworkAnalysis
PhishGuard: Multi-Faceted Phishing Detection: Leveraging URLs, HTML Features, and Visual Cues
Under Review in a Internation Journal of Information Technology
#MaskR-CNN
#FeatureExtraction
#DL
Unsupervised Multi-scale Segmentation of Cellular cryo-electron Tomograms with Stable Diffusion Foundation Model
Preprint: BioRxiv
#Stable Diffusion
#Unsupervised Learning
#LoRA
Academic Projects.
Following projects showcase my skills and experience through real-world examples of my work. Each project is briefly described with links to code repositories and live demos in it. It reflects my ability to solve complex problems, work with different technologies, and manage projects effectively.
FashionFushion Try-On
It enhances online shopping by enabling users to visualize clothing on themselves and receive personalized outfit recommendations using trend analysis and human-in-the-loop reinforcement learning. Developed for Myntra's HackerRamp WeForShe Hackathon, this project won pre-finalist status and I recieved a six-months internship opportunity.
#SegMind-Diffusion
#ChromaDB
#React JS
Legal Ninja
Client-Attorney Matchmaking Platform is an MLOps-driven solution designed to connect clients with legal experts efficiently. This platform utilizes advanced machine learning operations to create precise matches based on clients' specific legal needs and attorneys' specialized skills. We won MLOps Track in DataHack 2.0 Datathon.
#FalconAI
#Yake
#BERT
Carb-pool
AI-driven carpooling platform that connects commuters to reduce traffic congestion and carbon footprints. Designed to enhance urban travel, it features user matching, fare calculations, route optimization, and safety measures, incentivizing eco-friendly commuting with carbon credit rewards. We came second runner-up in Technovate 2023.
#DJango
#ARIMA
#TravellingSalesman
HungerZero
The main role of HungerZero is connect donner and NGO's, and also give a assistance to those who really need food, reduce wastage of food as well as hunger problem, donner can easily make a donation without any time wasting food shift will guide you throughout your process. We won Code Odyssey 2.0, 2023 with this project
#ReactJS
#DJango
#LogMealAPI
MindScape
It is designed for suicide detection and mental health monitoring, enabling timely intervention and support for at-risk individuals. The platform features a 24-hour chatbot for mental health assistance, stress relief games, mood-based music playlists, and guided meditation exercises, all aimed at enhancing user well-being.
#ReactJS
#uAgents
#Flask
ShopGPT
Shopping Chatbot with Web Scraping is an innovative solution that simplifies the online shopping experience by utilizing web scraping techniques to gather product information from various e-commerce platforms. Powered by Natural Language Processing, this helps users find the best deals through natural language interactions.
#Selenium
#Flask
#NLP
TensionNews
Event Tracker System tackles the challenge of managing information overload by developing a real-time news extraction platform. This system utilizes NLP techniques, large language models, and network analysis to categorize significant events and create an interactive network graph to visualize connections.
#NextJS
#Supabase
#VectorDB
Achievements.
DataHack 1.0 Winner, 2023
Rs.50,000
Thank you for visiting
