cv

Grader/Teaching Assistant

The University of Texas at Dallas

Teaching Assistant for CS3345, advised by Dr. Sridhar Alagar.

• Data Structures and Algorithms
  • Automate code grading to save 50% time, along with grading assignments, and exams for CS3345 - Data Structures and Algorithms, a class of 140 students.
  • Mentor students for improved understanding during office hours.

Data Science (NLP) Intern

ISN

Prospect Client Analytics powered by Intelligent Name Matching, advised by Smitha Jose.

• Deep Learning and Natural Language Processing
  • Innovated LSTM-based Intelligent Fuzzy Matching, a multifaceted solution for company matching, grouping, and industry prediction that enhanced over 3 production projects. Achieved 98% keyword prioritization accuracy and improved matching precision from 80% to over 90%. Developed in TensorFlow, utilizing Attention layers and k-fold TFIDF and accelerated by Trie and NumPy.
• Data Science
  • Engineered an end-to-end Prospect Client Analysis workflow that reduced prospecting time by 75%, and cut costs by 50%. Built in Alteryx with Python multiprocessing, utilizing NER for data cleaning from SQL client leads, coupled with Intelligent fuzzy matching.
  • Developed an automated solution to assess Cybersecurity Breaches involving contractor clients, using Selenium to web-scrape data for companies from Washington and Oregon.

AI Researcher

Intelligent Robotics and Vision Lab, UTD

Efficient Long Horizon Robot Exploration, Advised by Dr. Yu Xiang.

• Robot Exploration
  • Building a time-efficient Robot Exploration algorithm to achieve 95%+ 2D SLAM map coverage in unseen environments, utilizing Vision-Language Zero Shot Detection, Graph Neural Nets, Actor-Critic, and tools such as NVIDIA Isaac ROS and Stable Baselines3.
• Robot Navigation
  • Developing a SLAM-agnostic 2D robot PointNav algorithm using Soft Actor-Critic to efficiently guide the agent to its goal via the shortest path in a house environment.
• Tool Development
  • Engineered an interface in Python, facilitating over 3 lab projects, to set up a Robot Navigation environment, and perform controller and sensor simulation on Fetch robot in NVIDIA Isaac Sim.

Machine Learning Engineer

Quantrium AI

Leadership, Management and Development of AI Products, advised by Vishnu Kapalavai.

• Computer Vision
  • Led a team of 3 engineers to revamp a Smart Shelf Auditor to achieve 95% accuracy, and cut costs by 10%, by tuning a RetinaNet model, using GCP Linux servers. Improved the product sequence identifier accuracy by 15% by leveraging FRCNN Object Detection and OpenCV.
  • Researched ByteTrack - a Multiple Object Tracking Algorithm to build a Minimum Viable Product (MVP) on Sports Analysis for Soccer providing 5+ actionable game analytics utilizing PyTorch. This POC inspired the development of a new Video Analytics product for the company.
• Natural Language Processing
  • Pioneered a Document Sender Identifier app with 98.3% accuracy and boosted the process speed by 80%. Leveraged Spacy 3.2 NER with Transformer BERT embeddings to employ the smart keyword-matching logic powered by Python multiprocessing.
  • Conceptualized and developed a Proof of Concept (POC), successfully pitched to 2 prospective clients. Led the Minimum Marketable Product (MMP) development for a Bank Statements Analytics Project, guiding a team of 4 interns in building a pipeline using Kafka, Information Extraction, OCR, and advanced Statement Analytics using Topic modeling.
• Software Development
  • Built and led the Performance Monitoring of a Document Processor AI, achieving a boost of 90% in the process speed by leveraging Pandas Vectorization and Python Multiprocessing.
  • Developed an automated AWS Workspace troubleshooter that saved over 100 hours per month across the company.
  • Ensured consistent maintenance using CI/CD and dependency upgrades to the code repositories.
• Leadership
  • Initiated and led the development of new recruiting standards after interviewing over 50 ML and 30 SDE candidates.
  • Initiated best Software Development standards to achieve a 25% time reduction in code handovers.
  • Spearheaded the company's first semi-formal offline meetup, coordinating logistics and communication for approximately 15 employees to ensure a successful event.

Artificial Intelligence Engineer

Mirrag AI

Development of Computer Vision Products.

• Computer Vision
  • Developed a 90% accurate real-time License Plate Recognizer, harnessing custom YOLOv5 and PaddleOCR on Google Colab.
  • Engineered a 90% accurate Early-stage Fire Detector with over 30 FPS utilizing YOLOv5 which outperformed the Smoke Detector by 20 seconds.
  • Established a streamlined modular deployment pipeline for the applications, integrating tools such as PostgreSQL, imagezmq, and G-streamer for seamless operation.

Data Science Intern

Express Analytics

Development of NLP Products, advised by Vinay Dabhade.

• Natural Language Processing
  • Engineered a 93% accurate Sentiment API with emoji support for customer review analysis using Flair, Flask, and AWS Lambda.
• Data Science
  • Researched Fake Review Detection algorithm and proposed a multi-feature approach based on the User Profile information for enhanced accuracy.
  • Played a key role in the development of the Voice of Customer Analytics Dashboard, leveraging Seaborn and Pandas to provide actionable insights.

Segregation of Areca Nuts Using Three Band Photometry and Deep Neural Network

Soft Computing and its Engineering Applications, Springer, 2022

Designed and implemented a 90% accurate machine prototype for automated fruit sorting, utilizing a quantized YOLOv3 object detection model with 1 FPS on a Raspberry Pi 4 for high-precision quality classification, significantly improving agricultural processing efficiency.

DOG: Dynamic Object Grasping

Robotic grasping of object moving on 2D Deterministic Path with real-time Recursive Least Squares motion prediction.

Intelligent Name Matching

A BiLSTM-Attention model trained to prioritize keywords based on their Semantic role in a given Organization name.

GDBWat: Graph DB Watermarking

Implemented a secure and robust graph database watermarking method using pseudo-nodes to ensure data integrity and ownership verification.

IRVL Talk: NoMaD

A study and presentation talk on NoMaD: Goal Masked Diffusion Policies for Navigation and Exploration.

IRVL Talk: Real-world Mobile Manipulation with RL

A study and presentation talk on Continuously Improving Mobile Manipulation with Autonomous Real-World RL.

FetchNavIsaac: Robot Navigation Toolkit

An NVIDIA Isaac Sim Navigation starter kit with Sensor Simulation and Vision models.

InvestAID: AI powered Investment Dashboard

AI application that integrates ChatGPT 3.5, Google Stock Trends, and alternative data sources like news sentiments and social media trends, using advanced web scraping, sentiment analysis, topic classification, and few-shot meta-learning to predict Investible pointers.

Stair KeyDet: Keypoint Detection

A keypoint and person detection system for real-time staircase violation alerts using edge AI and camera feeds.

Master of Science

The University of Texas at Dallas

Computer Science - Intelligent Systems Track

Grade: 3.97

• Independent Research
• Artificial Intelligence
• Computer Vision
• Natural Language Processing
• Machine Learning
• Robot Manipulation and Navigation
• Statistics for Data Science

Bachelor of Technology

Indian Institute of Technology Dharwad

Mechanical Engineering, Minor in Machine Learning

Grade: 3.64

• Thesis Defense
• Deep Learning
• Machine Learning
• Robotics

Jonsson School Dean's Graduate Scholarship

Erik Jonsson School of Engineering and Computer Science, UTD

A merit-based scholarship with 50% tuition funding for Fall 2023 and Spring 2024 semesters.

Department Rank 2

Department of Mechanical Engineering, IITDh

Secured the rank among 36 students.