Pranav Harwadekar
Computer Science Honors Student at Texas A&M University (Minor in AI & Statistics). Passionate about building LLM-powered agents, scalable MLOps, autonomous navigation, and intelligent software systems.
About Me
Innovator, Software Engineer, and Machine Learning Researcher
Pranav Harwadekar
Location: Dallas, TX
University: Texas A&M University
Degree: B.S. CS, Minor in AI & Statistics
GPA: 3.93 / 4.0
Honors: President's Endowed Scholar & Engineering Honors
Email: pharwadekar@tamu.edu
Phone: (469) 426-8230
Hi, I'm Pranav! π
I am a Computer Science Honors student at Texas A&M University (minoring in Artificial Intelligence and Statistics) maintaining a 3.93 / 4.0 GPA as a President's Endowed Scholar.
My experience spans building production ML & software systems at top engineering organizations. At Bumble Inc., I developed LLM-powered agents integrated with Vertex AI, DBT, Airflow, and Snowflake to automate feature store validation and ML pipeline workflows across 40+ TB of production data. At McKesson Corporation, I designed PySpark demand forecasting and reorder point algorithms on Databricks.
I serve as the Software Engineer & Autonomous Navigation Engineer Lead at Karura USA, leading algorithm development in C++ and Python using ROS-based Nav2 and Stereo SLAM for URC rover competitions. I also led an AI predictive modeling initiative for American Airlines on 3M+ flight datasets, conducted data analytics externships with Beats by Dre (Apple) and Webacy Inc., and conducted ML research on hyperspectral algorithms and CUDA parallel processing at Texas A&M.
Beyond tech, I am deeply committed to community advocacy, having served two years as a National Youth Representative for UNICEF USA and currently working as a Cengage Student Ambassador.
Education
Academic Background & Scholastic Achievements
Texas A&M University
Bachelor of Science in Computer Science
Minor in Artificial Intelligence & Statistics
August 2023 β May 2027
Honors & Scholarships
- President's Endowed Scholar
- Engineering Honors Student
- Dean's Honor Roll
Relevant Coursework
Work & Leadership Experience
Software Engineering, MLOps, Autonomous Systems & Data Science
- Developing LLM-powered agents integrated with Vertex AI, DBT, Airflow, and Snowflake to automate feature store validation and machine learning pipeline workflows across 40+ TB of production data.
- Designing agent evaluation, testing, and guardrail frameworks to validate feature definitions, data lineage, and YAML-based ML artifacts, improving reliability of production ML infrastructure.
- Collaborating with Recommendation Systems and Trust & Safety ML teams to build scalable MLOps solutions that accelerate feature engineering, workflow orchestration, and model deployment.
- Designed and implemented reorder point algorithms and drug dispense forecasting algorithms using Apache Spark (PySpark) and MLflow on Databricks to optimize drug demand prediction and reorder timing.
- Leveraged spatial clustering and seasonal drug patterns to enhance prediction performance across retail and health system chains.
- Enhanced model performance by automating feature engineering and applying quantile-based forecasting to minimize inventory inefficiencies.
- Developing autonomous navigation algorithms in C++/Python for URC competition using ROS-based Nav2 and Stereo SLAM with RealSense D435 cameras, optimizing GPS waypoint following and sensor fusion (EKF) for precise localization across varied terrains.
- Simulating navigation and terrain algorithms in Gazebo and implementing image detection for ArUco markers.
- Boosted AI navigation accuracy by 25% through sensor data analysis and algorithm tuning with offshore Japanese team using Python and ROS.
- Developed AI predictive models for baggage arrival timing on over 3 million flights from Snowflakeβs Data Cloud, utilizing decision trees to capture complex relationships within flight operations.
- Analyzed model outputs to optimize ground operations, reducing passenger wait times to under 15 minutes for American Airlines.
- Conducted sentiment analysis on 1,200 Amazon reviews across 10 brands of headphones using manual and API scraping.
- Chosen among over 110 externs to present directly with the Head of Consumer Insights at Beats.
- Performed EDA analysis with Pandas, SciPy, and Seaborn and advanced NLP analysis with Gemini API for feature specific insights.
- Conducted rigorous Exploratory Data Analysis on smart contract risk tags analyzing over 965 lines of data across 270 contracts from 3 distinct databases and presented visualization data from SQL, Excel, Tableau.
- Profiled risks into 1 of 30 bins with unsupervised ML techniques (hierarchical clustering and Jaccard distance) for cluster analysis.
- Implemented hyperspectral algorithms in C++, including Jeffries Matusita Distance (JM), Spectral Angle Mapper (SAM), to enable precise spectral signature identification from complex hyperspectral data. Used WSL and Docker for cross-platform development.
- Created a C++ GUI using Glade and used CUDA for parallel computing, cutting hyperspectral data processing time by 40%.
Featured Projects
Autonomous Systems, AI Applications & Full-Stack Development
Autonomous Vehicle Navigation Simulation
Developing and integrating advanced sensor systems and navigation algorithms using CARLA Simulator, enhancing autonomous vehicle decision-making through machine learning.
- Leverages TensorFlow Graphs & PyTorch Geometric for spatial model learning.
- Utilizes Jupyter Notebooks & Google Colab for dataset processing & training.
AI for Justice β Criminality
Developed a real-time expression abnormality detector using a React frontend with Flask and AWS (S3 and Transcribe) for backend intelligence.
- Integrated OpenCV to analyze speech patterns and eye movement for deception cues.
- Compared visual analytics with NLP-extracted testimony to assess credibility.
Club3 Study Buddy Application
Collaborative study application connecting students with real-time group study sessions based on location and course topics.
- Built real-time session discovery using Maps API and Socket Programming.
- Utilized MongoDB for profiles, Clerk for auth, and Google Analytics for user metrics.
Technical Skills
Languages, Frameworks, Machine Learning Tools & Infrastructure
Get In Touch
Feel free to reach out for software engineering roles, research inquiries, or project collaborations!
Prefer to send a direct message?
Send Email Message