Sakshi Karande

Data | AI/ML | Software Engineering

About me

I have a keen interest in automating tasks and integrating intelligence into systems. Experienced in developing high-accuracy predictive models and multi-agent systems. I enjoy contributing to initiatives that utilize distributed big data to understand functionalities that aid impactful organizational decision-making.

Amherst, Massachusetts
Get In Touch View Work
my-photo

Education

Masters in Computer Science

University of Massachusetts, Amherst

September 2025 - present

GPA: -/4.0

Coursework Pursuing: COMPSI589 Machine learning; DACSS610 Data Science; COMPSCI685 Advanced NLP

Bachelors of Engg. in Data Science & Artificial Intelligence

University of Mumbai

August 2021 - May 2025

GPA: 3.7/4.0

Coursework Completed: C++; Object Oriented Programming: Java; Database Management System; Data Warehousing & Mining; Big Data Analytics(R programming); Software Engineering; Agile Project Management; Distributed Computing; Recommendation Systems; Advanced Artificial Intelligence; AI for finance & banking

Experience

Junior Analyst - Artificial Intelligence

Axia

November 2024 - May 2025

  • Built an agentic SEO training assistant using CrewAI and PhiData to automatically generate and evaluate employee training content.
  • Designed and deployed modular RESTful APIs with FastAPI to orchestrate agent behavior, memory, and role-based logic, enabling scalable multi-agent task execution.
  • Engineered "Futurists," a Japanese horoscope platform leveraging Groq LLMs, LangChain, and LangGraph to convert symbolic Yin-Yang charts into structured prompts—delivering witty daily Do’s and Don’ts and boosting user engagement.
  • Applied advanced prompt & context engineering and iterative tuning to enhance semantic accuracy and tone alignment, significantly reducing LLM hallucinations in production outputs.

AI Engineer & Research Intern

Axia

August 2024 - November 2024

  • Analyzed fleet trajectory data using pandas and NumPy to identify resource bottlenecks; implemented graph-based routing algorithms (BFS, DFS) that are now deployed in production, improving logistics efficiency.
  • Developed a sentiment classification pipeline by scraping and preprocessing YouTube comments with NLTK and TF-IDF vectorization; trained a Logistic Regression model achieving 97% accuracy.
  • Built an interactive quotation generation tool using ReTool, embedding JavaScript logic for dynamic pricing, discount automation, and conditional formatting, streamlining internal sales operations.
  • Extracted and cleaned semi-structured data from restaurant menus (PDF/HTML) using web scraping and manual parsing; conducted EDA, and created reports representing visual insights with matplotlib to inform menu design strategy.

Data Science Intern

TechCryptors

June 2023 - August 2023

  • Analyzed semi-structured data for 10,000+ real-estate records to identify market trends for stakeholder decision-making.
  • Performed exploratory data analysis (EDA) and developed predictive models using pandas, matplotlib, and seaborn to uncover actionable insights.
  • Engineered a Random Forest classification model achieving 92% accuracy, and applied K-means clustering to segment properties into 5 distinct location clusters; developed Tableau visualizations to effectively communicate analytical findings.
  • Delivered a comprehensive report enabling data-driven decisions by highlighting pricing trends, location clusters, and property type patterns.

Android Development Intern

Klaere Technologies

December 2022 - February 2023

  • Developed the TrekVisor app in Android Studio using Java to enable trek booking, safety features, and incident response.
  • Integrated Google Maps API and Weather API for accurate navigation and real-time updates.
  • Implemented a user-centric UI in XML and added emergency contact features to boost incident responsiveness.
  • Enhanced user experience by significantly improving safety and incident response compared to previous systems.

Projects

Sovereign Gold Price Prognostication using Multi-head Attention Based Transformers

Deep learning-based model for financial time-series forecasting

Deep Learning Multi-head attention Transformer Time Series Forecasting
  • Developed a multi-head attention transformer model to deliver high accuracy and reliable predictions
  • Improved forecasting accuracy of gold price surpassing SARIMA, RNN, Bi-LSTM using a decade-long dataset (2013–2023)
  • Utilized TensorFlow, NumPy, pandas, and scikit-learn; applied MinMaxScaler and temporal sequencing for data preprocessing and trained with the Adam optimizer for optimal performance
  • Achieved 98.94% accuracy and 1.06% MAPE, outperforming traditional forecasting methods and demonstrating the effectiveness of modern deep learning for financial time-series analysis
Code Demo

System for acquiring participants feedback for an in-house hackathon

Built for Microsoft Learn Students Club, VCET Student Chapter

Full-Stack Development LangChain Huggingface DevOps
  • Developed an end-to-end systems to streamline feedback acquisition, analysis, and support chatbot during the event
  • Designed an interactive feedback platform using ReactJS and Tailwind CSS, integrated with Firebase and an LLM API for automated analysis and PDF reporting
  • Implemented API for AI-powered chatbot using FastAPI, Langchain, and Huggingface Spaces (Dockerized) for integration of real-time conversational chatbot into the event website
  • Enhanced the hackathon experience for 250+ participants by delivering actionable feedback insights and significantly reducing organizing workload
Code Demo

Reinforcement Learning-Based Quantitative Bond Market Strategies

Optimization solution for bond trading with improved returns

Reinforcement Learning Financial Modeling Risk Management
  • A reinforcement learning solution demonstrating superior portfolio performance and risk-adjusted returns
  • Evaluated performance with total rewards, portfolio value, and Sharpe ratio beyond to optimize bond trading strategies
  • Built a Deep Q-Network (DQN) using TensorFlow, NumPy, and pandas; integrated epsilon-greedy policy, experience replay, and Adam optimizer to improve learning and decision-making
  • Achieved an 80% portfolio score and 22% return improvement over baseline using advanced RL algorithm (PPO)
Code Demo

Automated Government Document Processing System

Built for Aspiring Technologies Pvt. Ltd.

OCR REST APIs FastAPI Backend Engineering
  • Built an automation solution for bulk extraction and validation of data from scanned, semi-structured government documents.
  • Developed a FastAPI backend to process and extract barcodes using PyTesseract OCR, ensuring sequential data integrity throughout the pipeline.
  • Delivered machine-readable JSON responses, containerized and deployed the service via Docker, and validated all endpoints using Postman.
  • Automated document workflows, replacing manual effort of over 200 people and significantly increasing accuracy, scalability, and processing speed.
Code

Virtual Assistant enabled E-commerce Web App

Built for TechCryptors

Speech Recognition Python Programming MariaDB php
  • Enhanced user engagement and accessibility on an e-commerce platform by enabling hands-free shopping.
  • Engineered a virtual assistant that enables users to perform product searches and add items to cart or wishlist using real-time voice commands.
  • Developed and integrated a Google Web Speech API-powered assistant with backend RESTful APIs (FastAPI) and MariaDB using SQLAlchemy.
  • Delivered a production-ready voice assistant that improved AI-driven product experiences, user engagement, and accessibility.
Code

UniAttend: Face Recognition based Attendance System

Computer Vision Flask Pandas
  • Developed a web-based system to automate and accurately record student attendance using facial recognition technology.
  • Built UniAttend using Flask, HTML, CSS, JavaScript, and key libraries including face_recognition, OpenCV, Pillow, Pandas, and NumPy—enabling instant attendance marking via photo upload or webcam, with persistent records and ~99% recognition accuracy.
  • Reduced manual effort and eliminated duplicate entries, enabling batch attendance for 70+ students in under 2 seconds per recognition event.
  • This project addresses the problem of traditional attendance tracking in educational institutions, which was manual, slow, and prone to errors.
Code

Technical Skills

Languages

Software Development

FastAPI
RESTful APIs
Postman
ReTool
Flask

Data & Analytics

pandas
NumPy
scikit-learn
matplotlib
seaborn
EDA
Time-Series Forecasting
Feature Engineering
Predictive Modeling
Statistical Analysis
Regression
Clustering
Tableau

AI & ML

Keras
LLMs (Groq, Gemini, Hugging Face)
LangChain
LangGraph
Context & Prompt Engineering
NLP
Reinforcement Learning
Recommendation Systems
CrewAI framework
Phidata framework

Deployment & Collaboration

Hugging Face Spaces
Google APIs (Maps, Speech)
Agile

Publications

Learning Efficacy Augmentation for Bose—Einstein Condensates of Numerical Representations: A Review

IEEE Computer Society - 2024 First International Conference on Data, Computation and Communication (ICDCC)

April 2025

Authors: Sakshi Karande; Tej More; Raunak Joshi; Neha Raut

View Publication

Sovereign Gold Price Prognostication using Multi-head Attention Based Transformers

IEEE - 3rd International Conference on Sustainable Computing and Smart Systems (ICSCSS 2025)

August 2025

Authors: Sakshi Karande; Tej More; Raunak Joshi; Neha Raut

Publication under progress

Unlocking the Power of Data Augmentation with Generative Adversarial Networks (GANs)

VCET Techzette - Technical Magazine [ISSIN:2584-0886]

February 2025

Author: Sakshi Karande

View Publication

Achievements

'Top student' in the third year of Bachelors of Engineering

2023

Department of Data Science & AI at VCET(University of Mumbai)

Winner

2023

Oscillation - Technical Paper Presentation Competition

Winner

2023

Paperathon - Technical Research Competition

'Runner up' at State Entrepreneurship Competition

2021

Issued by Vivek College of Science (University of Mumbai)

Microsoft Azure Certified: Cloud Fundamentals

2023

Certified by Microsoft Learn Students Club

Java Programming Certification

2023

Certified by Infosys Springboard

Extra-curricular

AI/ML Development Lead @Microsoft Learn Students Club - Student Chapter

Sept 2023 - May 2025

Orchestrated comprehensive Git & GitHub workshop for 250+ sophomore students, delivering hands-on training in version control fundamentals, collaborative development workflows, and industry-standard software development practices

Social Media & Technical Team Member @National Students Data Corps- Student Chapter

Aug 2022 - May 2025

Corporate & Public Relations @E-Cell VCET

Aug 2021 - Aug 2022

Contact Me

Get in Touch

Hey there! Have a project in mind or just want to connect? I'd love to hear from you! Drop me a message below and I'll get back to you soon!

Massachusetts, United States
skarande@umass.edu