⌘
Programming & Data
Python, data preprocessing, exploratory data analysis, dataset validation and structured data workflows.
Hello, It's Me
Harshita Ganti
And I'm an AI/ML Engineer|
From real-world data analytics to machine learning research, I love turning complex data into practical, meaningful solutions.
About Me
A blend of industry experience, research curiosity, and hands-on problem solving.
My story
Analyst experience. AI ambitions.
I am an M.Tech student specializing in Artificial Intelligence, with prior professional experience as a Content Data Analyst at ServiceNow. My interests span machine learning, deep learning, retrieval-augmented generation and applied analytics. I enjoy turning messy data into structured information and building systems that make research, prediction and decision-making more accessible. My portfolio brings together enterprise data work, research-oriented prototypes, computer vision models and collaborative hackathon experience.
I enjoy model experimentation, exploring the reasoning behind predictions, and presenting technical work clearly through measurable results.
Hyderabad, IndiaM.Tech AI/MLData AnalyticsApplied Machine Learning
My Skills
Not just a list of buzzwords — these are the areas I work with across analytics and projects.
◈
Machine Learning
Supervised learning, model evaluation, feature engineering, experimentation and predictive analytics.
✳
Deep Learning & AI
Deep learning concepts, temporal modeling, Generative AI fundamentals and applied AI research.
▦
Data Analytics
Data quality, analysis workflows, interpreting patterns and communicating findings.
⬡
Research & Evaluation
Literature review, baseline comparisons, experimental design and model-performance reporting.
↗
Team Projects
Collaborative ML development, team coordination, iterative improvement and presenting technical outcomes.
↗
Entity Resolution & Validation
Applied record normalization and entity-matching thinking in enterprise data work and the Amazon ML Challenge, including candidate generation and validation workflows.
▣
Model Evaluation
Experience assessing classification and prediction results, examining model outputs and comparing alternatives; current supplier-risk research remains in progress.
Professional Experience
Real enterprise data, real-world impact.
August 2023 — February 2025 · Hyderabad
Content Data Analyst · ServiceNow
Worked with large-scale network discovery data and software asset records to improve the consistency and reliability of information used in IT Asset Management (ITAM). My role combined detailed software research, normalization rules, content-library maintenance and knowledge documentation.
A glimpse into my enterprise data workflowWhere I contributed
Data Quality & Normalization
Standardized raw discovery records, researched software titles, versions and editions, and mapped ambiguous entries to accurate software identities.
IT Asset Management
Maintained structured software-library content supporting reliable asset identification and license-management workflows.
Research & Documentation
Investigated software publishers and product families, authored ITAM knowledge-base content, and improved the clarity of reference materials.
Skills developed through real-world data
Data NormalizationEntity MappingData QualitySoftware Asset ResearchKnowledge Management
Selected work · Real screenshots
Featured Projects
Detailed case studies with project goals, implementation notes, real screenshots and reported outcomes. Select a gallery image to inspect it at full size.
01 / M.Tech major project · In development
Pharmaceutical Supply Chain Resilience
Pharma Control Tower: a supplier-risk and procurement-intelligence dashboard. The current interface provides supplier risk segmentation, supplier profiles, and purchase-order visibility. Research on supplier efficiency and late-delivery prediction using DEA, TCN and comparative ML evaluation is ongoing; predictive performance has not yet been finalized.
Supplier riskProcurement analyticsDEA / TCN — researchIn progress
537suppliers in dashboard
5,467purchase orders
94high / critical risk
The problem
Procurement teams need a consolidated view of supplier reliability, order delays and emerging delivery risks. Fragmented supplier and purchase-order records can make prioritization difficult.
Current implementation
The Pharma Control Tower prototype consolidates supplier and order records into an overview dashboard, supplier-risk distribution, performance-versus-risk visualization and high-risk supplier profiles. The displayed dataset currently includes 537 suppliers and 5,467 purchase orders.
Research direction
The M.Tech work draws on supplier-efficiency analysis with Data Envelopment Analysis (DEA) and late-delivery modeling with Temporal Convolutional Networks (TCN). Comparative ML experiments and validation are planned or underway, not final results.
What the screenshots prove
The images show a functioning analytical interface and risk-profile views. Dashboard risk scores are operational outputs; they should not be interpreted as independently validated model accuracy.
02 / Generative AI · NLP
Intelligent Academic Research Assistant
A research assistant that processes academic PDFs, supports retrieval-augmented question answering, represents research concepts through knowledge graphs, and visualizes keyword trends. Explore the actual application screens below.
PythonRAGKnowledge GraphsNLPResearch Analytics
04application screens
RAGretrieval workflow
Purpose & workflow
Designed to help students navigate large volumes of academic literature. Users can upload research PDFs, process document text and explore paper content using AI-assisted retrieval and research-focused interfaces.
RAG & knowledge representation
The project combines retrieval-augmented generation with knowledge-graph representations to organize relationships among papers, topics and concepts. This provides a structured complement to text-based search and generated answers.
Features demonstrated
The gallery shows successful PDF processing, a visual graph of connected research concepts and a keyword-frequency dashboard for identifying recurring themes.
Engineering focus
Document ingestion, semantic retrieval, concept extraction, visualization and usable research workflows. The screenshots document features, not benchmarked answer-quality scores.
03 / Hackathon · 2026
Amazon ML Challenge
Collaborative machine-learning work on structured-data entity matching. Our team achieved a 0.86 leaderboard score. The photograph below captures my actual development workflow.
PythonEntity MatchingModel validationTeam project
0.86leaderboard score
Challenge & approach
Worked with a team on matching related entities across structured data sources. The workflow involved analyzing source datasets, generating candidate matches, applying normalization and matching logic, and validating the resulting predictions.
My contribution
Contributed to the team's development and testing process, including code for dataset analysis, candidate generation and strict validation. The accompanying photograph documents the implementation environment.
Outcome
Our team achieved a 0.86 leaderboard score. This is the reported challenge leaderboard result, not a claim of production accuracy or a final competition rank.
Behind the build · Actual coding session
Brain Tumor Detection in MRI
EfficientNet and CNN transfer learning with image preprocessing and augmentation; reported 95%+ accuracy in the project.
EfficientNetCNN95%+ reported
Implementation details
Prepared MRI images through resizing, normalization, noise reduction and augmentation before fine-tuning EfficientNet and CNN-based classifiers using transfer learning.
Reported result
The project reports accuracy above 95%. This is a project result and should not be interpreted as a clinical validation or diagnostic claim.
Real-Time Emotion Recognition
Facial expression classification with CNN, VGG16 and ResNet; reported up to 90% accuracy in the project.
VGG16ResNet90% reported
Implementation details
Built a facial-expression recognition pipeline using CNN architectures including VGG16 and ResNet. Trained and evaluated on image datasets with a focus on classification from images and video frames.
Reported result
Up to 90% accuracy was reported in the project; real-world performance depends on dataset conditions and deployment context.
Official credential verification
My Certifications
Clean credential cards with direct links to the official verification pages — no mobile screenshots.
◆
Databricks Academy · October 8, 2026Generative AI Fundamentals
Databricks Accredited Generative AI Fundamentals.
View verified credential ↗◈
Oracle · Verified badgeAgentic AI Certified Foundations Associate
Foundational knowledge of agentic AI concepts and systems.
View verified credential ↗✧
Additional trainingContinuous Learning
Software Developer Internship — Exposys Data Labs; Complete Python Bootcamp — Udemy; Azure Cloud Computing — Verzeo.
Training listed in my resume; verification links not supplied.
Academic journey
My Education
M.Tech — Artificial Intelligence
Mahatma Gandhi Institute of Technology, Hyderabad
B.Tech — Information Technology
Guru Nanak Institutions Technical Campus
Let's Connect
Open to AI/ML engineering, data science and GenAI opportunities. Reach out via LinkedIn or email.