About
I am an AI Autonomy Intern at Skydio, building AI systems for the Autonomy team, and pursuing an MS in Computer Science at the University of Massachusetts Amherst (GPA 4.0/4.0). My experience spans agentic AI, applied computer vision, NLP, and ML platform engineering. I focus on turning research-driven models into practical production systems.
Experience
Travelers
Data Science Intern, DSLDP
- Built a 3-tier pipeline to flag auto-insurance pricing disparities: a deterministic engine flags variables, a grounded LLM generates narratives, and an agentic RAG chatbot with tool calling aids reasoning and report generation
- Delivered the solution as a Streamlit dashboard spanning coverages and quarters; scaled to 20+ concurrent users on a modular plug-in architecture that enables onboarding without disrupting production workflows
- Cut quarterly back-test review from 5 hours to 20 minutes at $0.05 per query in LLM inference cost, with 100% recall against analyst labels, enabling faster model refits and lower-latency pricing decisions for state teams
Ascend Cargo Systems
AI/ML Engineer Intern
- Architected a production-grade agentic AI system with a custom Java Spring Boot MCP server and dual-LLM design, enabling real-time freight retrieval and autonomous carrier assignment across Slack, Microsoft Teams, email, and web chat, reducing manual assignment by 80%
- Built an asynchronous document-processing system leveraging GPT-4o Vision and Microsoft Graph API to automate proof-of-delivery verification, achieving 10× higher throughput and reducing latency from 3 hours to 10 min
Cogniverse Labs
AI/ML Intern, Product Engineering
- Fine-tuned Faster R-CNN and YOLOv8 models to detect roadside garbage, footpath encroachments, and other urban infrastructure issues for a government Smart City project
- Deployed the computer vision models using TorchServe, reducing manual inspection workload by 40%
- Focused on minimising false positives in real-world street imagery and collaborated with the product engineering team to translate model development into practical urban-monitoring solutions
National University of Singapore
Machine Learning Intern & Teaching Assistant
- Developed a medical CV pipeline using Vision Transformer with progressive unfreezing and domain-specific augmentation achieving 96% diagnostic accuracy; deployed real-time inference via a Flask REST API
- Architected and deployed an SVM-based churn prediction microservice for a Tier-2 telecom provider by engineering high-dimensional customer feature sets to achieve 92% accuracy
- Served as a Teaching Assistant for the AIYA program in summer 2024, guiding around 45 students from India and the Middle East under Dr. Sanka Rasnayaka at NUS School of Computing
- Supported hands-on ML model deployment with AWS SageMaker and chatbot development with AWS Lex while mentoring students through their project work
Colt Technology Services
Data Science Intern
- Automated sentiment extraction from employee feedback, by designing and training a BERT-based transformer model with end-to-end NLP pipelines, reducing manual analysis effort and achieving 90% model accuracy
Projects
Multilingual Online Polarization Detection
GitHub- Engineered an XLM-RoBERTa framework for Polarization Detection across 9 languages using NLLB-200 for translation-based oversampling to resolve manifold collapse and increase Macro F1 from 0.63 to 0.80
- Integrated Supervised Contrastive Learning, Focal Loss and Dynamic Thresholding to disentangle overlapping rhetorical labels, achieving a further 8% boost in multi-label assignment
Distributed ML Pipeline for Flight Delay Prediction
GitHub- Built a distributed flight-delay prediction pipeline on 1M+ records in PySpark, leveraging the Catalyst optimizer and lazy evaluation to scale 4× faster than Pandas
- Tuned hyperparameters via grid search to achieve 0.88 AUC and validated scalability by benchmarking PySpark against Pandas
Publications
Diagnosis-based IQA and Enhancement for Low-Dose CT Images
View PaperPublished in Frontiers in Radiology, 2025 (PubMed Indexed)
- Authored a diagnosis-based Image Quality Assessment and enhancement technique for Low-Dose CT scans, leveraging feature extraction and spatial enhancement methodology
- The SVR-GS model achieved a 0.969 PLCC, with enhancements clinically verified by Dr. Inthulan Thiraviraj
ANGEL: Automated Navigation Giving Emergency Location — A Robust Tracking System
IEEE XplorePublished in 2025 International Conference on Information, Implementation, and Innovation in Technology (I2ITCON) — IEEE
Efficient Medical Diagnosis Using ML Algorithm: Comparative Study on Cataract Detection and Analysis
View PaperPublished in Advances in Electrical and Computer Technologies — ICAECT 2024 (CRC Press / Taylor & Francis, Scopus Indexed)
Certifications
AWS Certified Cloud Practitioner
VerifyFoundational cloud architecture, security, and AWS service knowledge.
AWS Certified Solutions Architect – Associate
VerifyDesign of scalable, secure, and cost-optimized cloud infrastructure.
Contact
Open to AI/ML engineering roles, internships, and research collaborations.