Sweta Singh
Package Consultant @IBM | Aspiring Data Scientist | ML, NLP, RAG, LLMs | Python, SQL | PG Diploma – Data Science & AI, IIIT Bangalore | Solving Real-World Problems with Data | Actively seeking Data Science opportunities
- Role
- Package Consultant at IBM
- Location
- Bengaluru, KA, IN
- LinkedIn followers
- 500 followers
About Sweta Singh
I am an AI & Data Science professional with hands-on experience in building NLP and GenAI…
Experience
Package Consultant
Mar 2023 — Present · Bengaluru, IN
IBM Intelligent Role Remediation Agent – AI-Driven Automated Ticket Resolution• Designed and deployed an Agentic AI workflow system on IBM Cloud using LangGraph, MCP, and containerized pods.• Implemented a multi-tenant architecture supporting image/text ingestion, auto-fulfilling low-impact tasks and routing complex issues to admins.• Integrated with ServiceNow ITSM using fulfilment APIs with robust retries, error-tolerance, and idempotent execution.• Built polling, callback, and notification pipelines enabling real-time tracking of workflow and ticket status.• Developed automated issue diagnosis, SAP/Oracle application classification, ambiguity detection, and duplicate-ticket prevention.• Delivered a secure, scalable production system with admin-configurable endpoints, audit trails, and alerting for failed Customer Churn Prediction – Predictive Analytics• Built churn prediction models using Logistic Regression and Random Forest, achieving ~85% accuracy.• Applied PCA, feature engineering, SMOTE imbalance handling, and interpretability techniques including feature importance & SHAP.• Generated actionable insights enabling business teams to prioritize high-risk customer segments for retention strategies.• Automated end-to-end data processing, model training, and comparison pipelines.• Delivered dashboard-ready outputs showing churn probability, key drivers, and customer risk Querying System with Local LLMs – Secure Document Q&A• Developed a secure PDF-based Q&A engine using LangChain + local Flan-T5, enabling fully offline inference.• Implemented context-aware chunking, FAISS vector search, and RAG for semantic retrieval.• Used PyMuPDF for accurate PDF parsing, metadata extraction, and document cleanup.• Built a modular embedding pipeline supporting Word2Vec, BERT, and Instructor embeddings without code refactoring.
Find verified contacts for anyone on LinkedIn
Unifers gives sales teams verified emails and direct dials, enriched profiles, and outreach that lands in the inbox.
Free plan included · No credit card required
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.