Claims you can check
Each card points at something real — open the source and verify for yourself.
- StatedRésuméBuilt a multimodal document-understanding pipeline using open-source vision-language models to extract text, tables, charts, signatures, and layout from multi-page financial documents, replacing traditional OCR-based extraction. Orchestrated the multi-stage pipeline using LangGraph, built a CrewAI-based validation agent using the ReAct pattern to verify extracted fields, and structured outputs into machine-readable JSON.
- StatedRésumé
AI Hallucination Detection System
Built an LLM validation pipeline performing claim extraction, web evidence retrieval, and semantic similarity scoring to detect hallucinations. Integrated Google Gemini API and FAISS vector search for real-time evidence grounding. Designed an LLM observability layer providing trust scoring and reliability monitoring for AI-generated outputs, deployed via FastAPI with an interactive Streamlit dashboard. - StatedRésumé
AskQL - Natural Language to SQL Query System
Built an AI-powered natural language-to-SQL system converting plain English queries into structured SQL commands using rule-based and NLP-driven pipelines. Applied classical NLP techniques for query understanding, with exploratory LLM-based enhancements and a modular architecture for future generative AI integration. - StatedRésumé
Multimodal Financial Document Intelligence
Financial institutions process large volumes of dense, multi-page documents. Used open-source vision-language models (LLaVA, Qwen-VL) to jointly read text, tables, charts, signatures and layout directly from page images. Orchestrated extraction, validation, and structuring as a LangGraph pipeline with a CrewAI validation agent. - StatedRésumé
AI Hallucination Detector
Claim extraction breaks an LLM response into discrete assertions. FAISS vector search retrieves relevant evidence, and cosine similarity between embeddings produces a trust score. Deployed as a FastAPI service with a Streamlit dashboard. - StatedRésumé
Industrial RAG + Anomaly Intelligence
Built a RAG system over equipment logs and SOPs. Built a companion anomaly-detection pipeline across 1,400+ sensors and linked flagged anomalies to the retrieval system. Evaluated using RAGAS against a 100-observation golden dataset. - StatedRésumé
GPT from Scratch - PyTorch
Implementation of tokenization, attention, training dynamics, and autoregressive generation built from scratch. - StatedRésumé
System Design RAG Assistant
Conversational retrieval assistant for system design and software engineering books to avoid ungrounded LLM responses. - StatedRésumé
AI MOM - Meeting Minutes Generator
Tool for transcribing meetings and creating structured minutes, decisions, and action items automatically. - StatedRésumé
AskQL - Natural Language to SQL
Natural language to SQL interface helping business users query data without SQL knowledge. - StatedRésumé
ASR Shootout - Indian Conversational Speech
Evaluation framework for ASR models on noisy, conversational, multilingual speech and named entities. - StatedRésumé
Fraud Detection MLOps Platform
MLOps platform for fraud models encompassing reproducibility, experiment tracking, deployment, monitoring, and inference. - StatedRésumé
Fake News Trend Drift Detector
Model monitoring and drift detection system to detect silent model degradation and trigger retraining. - StatedRésumé
Synthetic Data Generation Platform
Platform for generating synthetic data while preserving statistical relationships and maintaining privacy. - StatedRésumé
AI-Powered Investment Portfolio Platform
Integrated tool for retail investors providing live portfolio tracking, market analysis, alerts, and insights. - StatedRésumé
AI Symptom Checker
Clinical decision support tool assisting users in interpreting symptoms before consulting professionals. - StatedRésumé
Health Insurance Fraud Detection System
Machine learning system identifying subtle fraud patterns across claims, providers, patients, and transaction attributes. - StatedRésumé
Structural Defect Analyzer
Computer vision application for inspecting infrastructure and identifying visible cracks and defects across large structures. - StatedRésumé
Credit Card Fraud Detection System
Financial risk analytics project dealing with highly imbalanced transaction data for fraud detection. - StatedRésumé
AI Resume Skill Matcher
NLP tool comparing resume skills with job requirements beyond simple keyword matching. - StatedRésumé
Data Quality & Pipeline Monitoring Dashboard
Monitoring dashboard providing visibility into incoming data health and running data quality checks. - StatedRésumé
Synthetic Data Generation Platform
Generates synthetic data preserving statistical relationships for privacy-constrained datasets. - StatedRésumé
AI-Powered Investment Portfolio Platform
Platform for live portfolio tracking, market analysis, alerts, and portfolio-level insights. - StatedRésumé
AI Symptom Checker
Assists users in interpreting symptoms and determining appropriate medical attention levels. - StatedRésumé
Health Insurance Fraud Detection System
Identifies subtle fraud patterns across claims, providers, patients, and transaction attributes. - StatedRésumé
Multimodal Financial Document Intelligence
Used open-source vision-language models to process dense multi-page financial documents, orchestrated as a LangGraph pipeline with a CrewAI validation agent. - StatedRésumé
AI Hallucination Detector
Breaks LLM response into assertions, uses FAISS vector search to retrieve reference evidence, and computes trust scores. Deployed with FastAPI and Streamlit. - StatedRésumé
Industrial RAG + Anomaly Intelligence
Built a RAG system over 1,000+ equipment logs and SOPs paired with an anomaly detection pipeline across 1,400+ SCADA/BMS sensors. - StatedRésumé
GPT from Scratch - PyTorch
Building a GPT model from first principles including tokenization, attention, training dynamics, and autoregressive generation. - StatedRésumé
System Design RAG Assistant
RAG assistant for searching large system-design and software-engineering books conversationally. - StatedRésumé
AI MOM - Meeting Minutes Generator
Automated transcribing meetings and creating structured minutes, decisions, and action items. - StatedRésumé
AskQL - Natural Language to SQL
Translates natural language questions into SQL queries for business users. - StatedRésumé
ASR Shootout - Indian Conversational Speech
Evaluates ASR models on noisy, conversational, multilingual speech and named entities. - StatedRésumé
Fraud Detection MLOps Platform
MLOps platform for fraud models covering reproducibility, experiment tracking, deployment, monitoring, and inference. - StatedRésumé
Fake News Trend Drift Detector
Monitors deployed models for production data drift and triggers retraining. - StatedRésumé
Structural Defect Analyzer
Automates infrastructure inspection by detecting cracks and visible defects across large structures. - StatedRésumé
Credit Card Fraud Detection System
Fraud detection system utilizing sampling techniques to address extreme class imbalance. - StatedRésumé
AI Resume Skill Matcher
Compares resume skills with job requirements using semantic similarity scoring. - StatedRésumé
Data Quality & Pipeline Monitoring Dashboard
Monitors health, completeness, and structure of incoming datasets across data pipelines. - StatedRésumé
Trainee – Manufacturing & Formulation @ CIPLA Ltd
Developed GenAI, RAG and anomaly-detection systems supporting plant operations alongside official manufacturing responsibilities. - StatedRésumé
Trainee – Manufacturing & Formulation @ CIPLA Ltd
Official role in Manufacturing & Formulation. Developed GenAI/RAG and sensor anomaly-detection systems supporting plant operations. Built RAG document processing over 1,000+ logs/SOPs, processed sensor time-series data from 1,400+ sensors across 15 SCADA and 47 BMS systems, and performed LLM evaluation using RAGAS. - StatedRésumé
Data Scientist @ CIPLA Ltd
Built a RAG-based document processing system indexing 1,000+ equipment logs and SOPs, reducing monthly troubleshooting job cards from 120+ to ~40 (~67% reduction). Applied prompt engineering to tune retrieval and response quality across multiple LLM-backed workflows. Evaluated RAG pipeline outputs periodically against a 100-observation golden dataset using RAGAS metrics to evaluate and improve model performance. Built an anomaly detection pipeline across 1,400+ sensors, integrated with a knowledge retrieval system to cut root-cause analysis time by 55–60%. - StatedRésumé
Performer of the Month
Recognized as Performer of the Month three consecutive times at CIPLA Ltd for project delivery on RAG and anomaly-detection work. - StatedRésumé
Performer of the Month
Recognized as 'Performer of the Month' three consecutive times at CIPLA Ltd for consistent delivery on RAG and anomaly-detection work.
What Shreya can prove
Proven Unproven
- agentic AI: self-declared, 0 evidence items
- C: self-declared, 0 evidence items
- CNN: self-declared, 1 evidence items, linked to Project: Structural Defect Analyzer
- CrewAI: self-declared, 2 evidence items, linked to Project: Multimodal Financial Document Understanding System, Project: Multimodal Financial Document Intelligence
- CTGAN: self-declared, 1 evidence items, linked to Project: Synthetic Data Generation Platform
- Deep Learning: self-declared, 0 evidence items
- Deepgram: self-declared, 1 evidence items, linked to Project: ASR Shootout - Indian Conversational Speech
- Docker: self-declared, 2 evidence items, linked to Project: Fraud Detection MLOps Platform, Project: Fake News Trend Drift Detector
- DVC: self-declared, 1 evidence items, linked to Project: Fake News Trend Drift Detector
- Evidently AI: self-declared, 1 evidence items, linked to Project: Fake News Trend Drift Detector
- FAISS: self-declared, 2 evidence items, linked to Project: AI Hallucination Detection System, Project: AI Hallucination Detector
- FastAPI: self-declared, 6 evidence items, linked to Project: AI Hallucination Detection System, Project: AskQL - Natural Language to SQL Query System, Project: AI Hallucination Detector, Project: AI MOM - Meeting Minutes Generator, Project: AskQL - Natural Language to SQL, Project: Fraud Detection MLOps Platform
- Feature Engineering: self-declared, 0 evidence items
- Flask: self-declared, 0 evidence items
- Generative AI: self-declared, 1 evidence items, linked to Project: AskQL - Natural Language to SQL Query System
- Git: self-declared, 0 evidence items
- Git / GitHub: self-declared, 0 evidence items
- GitHub: self-declared, 0 evidence items
- Google Gemini API: self-declared, 2 evidence items, linked to Project: AI Hallucination Detection System, Project: AI Hallucination Detector
- Google STT: self-declared, 1 evidence items, linked to Project: ASR Shootout - Indian Conversational Speech
- Gradient Boosting: self-declared, 1 evidence items, linked to Project: AI Symptom Checker
- Grafana: self-declared, 1 evidence items, linked to Project: Fake News Trend Drift Detector
- Hugging Face Transformers: self-declared, 0 evidence items
- Java: self-declared, 0 evidence items
- JavaScript: self-declared, 0 evidence items
- JiWER: self-declared, 1 evidence items, linked to Project: ASR Shootout - Indian Conversational Speech
- Jupyter Notebook: self-declared, 0 evidence items
- JWT: self-declared, 1 evidence items, linked to Project: AI-Powered Investment Portfolio Platform
- Keras: self-declared, 1 evidence items, linked to Project: Structural Defect Analyzer
- LangChain: self-declared, 1 evidence items, linked to Project: Industrial RAG + Anomaly Intelligence
- LangGraph: self-declared, 2 evidence items, linked to Project: Multimodal Financial Document Understanding System, Project: Multimodal Financial Document Intelligence
- LightGBM: self-declared, 1 evidence items, linked to Project: Fraud Detection MLOps Platform
- LLaVA: self-declared, 2 evidence items, linked to Project: Multimodal Financial Document Understanding System, Project: Multimodal Financial Document Intelligence
- LLM Fine-tuning: self-declared, 0 evidence items
- LLM systems: self-declared, 0 evidence items
- LLMs: self-declared, 1 evidence items, linked to Project: AI MOM - Meeting Minutes Generator
- Matplotlib: self-declared, 1 evidence items, linked to Project: Credit Card Fraud Detection System
- MLflow: self-declared, 2 evidence items, linked to Project: Fraud Detection MLOps Platform, Project: Fake News Trend Drift Detector
- Multimodal LLMs: self-declared, 1 evidence items, linked to Project: Multimodal Financial Document Understanding System
- MySQL: self-declared, 0 evidence items
- Next.js: self-declared, 1 evidence items, linked to Project: AI-Powered Investment Portfolio Platform
- NumPy: self-declared, 0 evidence items
- OpenCV: self-declared, 1 evidence items, linked to Project: Structural Defect Analyzer
- Pandas: self-declared, 5 evidence items, linked to Project: ASR Shootout - Indian Conversational Speech, Project: Synthetic Data Generation Platform, Project: Health Insurance Fraud Detection System, Project: Credit Card Fraud Detection System, Project: Data Quality & Pipeline Monitoring Dashboard
- Pandas / NumPy: self-declared, 0 evidence items
- pgvector: self-declared, 1 evidence items, linked to Project: Industrial RAG + Anomaly Intelligence
- PostgreSQL: self-declared, 1 evidence items, linked to Project: AI-Powered Investment Portfolio Platform
- Power BI: self-declared, 1 evidence items, linked to Project: Data Quality & Pipeline Monitoring Dashboard
- Prefect: self-declared, 1 evidence items, linked to Project: Fake News Trend Drift Detector
- Prompt Engineering: self-declared, 0 evidence items
- Prompt Tuning: self-declared, 0 evidence items
- Python: self-declared, 6 evidence items, linked to Project: System Design RAG Assistant, Project: AI MOM - Meeting Minutes Generator, Project: AskQL - Natural Language to SQL, Project: Fraud Detection MLOps Platform, Project: AI Symptom Checker, Project: Health Insurance Fraud Detection System
- PyTorch: self-declared, 1 evidence items, linked to Project: GPT from Scratch - PyTorch
- Qdrant: self-declared, 1 evidence items, linked to Project: System Design RAG Assistant
- Qwen-VL: self-declared, 2 evidence items, linked to Project: Multimodal Financial Document Understanding System, Project: Multimodal Financial Document Intelligence
- RAG: self-declared, 3 evidence items, linked to Project: AI Hallucination Detection System, Project: Industrial RAG + Anomaly Intelligence, Project: System Design RAG Assistant
- RAG pipelines: self-declared, 0 evidence items
- RAGAS: self-declared, 1 evidence items, linked to Project: Industrial RAG + Anomaly Intelligence
- Random Forest: self-declared, 1 evidence items, linked to Project: AI Symptom Checker
- ReAct: self-declared, 4 evidence items, linked to Project: Multimodal Financial Document Understanding System, Project: AskQL - Natural Language to SQL Query System, Project: Multimodal Financial Document Intelligence, Project: AskQL - Natural Language to SQL
- REST APIs: self-declared, 0 evidence items
- SCADA/BMS Data: self-declared, 1 evidence items, linked to Project: Industrial RAG + Anomaly Intelligence
- Scikit-learn: self-declared, 4 evidence items, linked to Project: Synthetic Data Generation Platform, Project: AI Symptom Checker, Project: Health Insurance Fraud Detection System, Project: Credit Card Fraud Detection System
- SDV: self-declared, 1 evidence items, linked to Project: Synthetic Data Generation Platform
- Semantic Similarity Scoring: self-declared, 2 evidence items, linked to Project: AI Hallucination Detection System, Project: AI Resume Skill Matcher
- Sentence Transformers: self-declared, 3 evidence items, linked to Project: AI Hallucination Detector, Project: System Design RAG Assistant, Project: AI Resume Skill Matcher
- Small Language Models (SLMs): self-declared, 0 evidence items
- SMOTE: self-declared, 1 evidence items, linked to Project: Credit Card Fraud Detection System
- spaCy: self-declared, 1 evidence items, linked to Project: AskQL - Natural Language to SQL
- SQL: self-declared, 3 evidence items, linked to Project: AskQL - Natural Language to SQL Query System, Project: AskQL - Natural Language to SQL, Project: Data Quality & Pipeline Monitoring Dashboard
- SQLite: self-declared, 0 evidence items
- Statistical Analysis: self-declared, 0 evidence items
- Streamlit: self-declared, 2 evidence items, linked to Project: AI Hallucination Detection System, Project: AI Hallucination Detector
- TensorFlow: self-declared, 1 evidence items, linked to Project: Structural Defect Analyzer
- TF-IDF: self-declared, 1 evidence items, linked to Project: AI Resume Skill Matcher
- Transformer: self-declared, 1 evidence items, linked to Project: GPT from Scratch - PyTorch
- TypeScript: self-declared, 1 evidence items, linked to Project: Synthetic Data Generation Platform
- Vector Databases: self-declared, 0 evidence items
- WebSockets: self-declared, 1 evidence items, linked to Project: AI-Powered Investment Portfolio Platform
- Whisper: self-declared, 2 evidence items, linked to Project: AI MOM - Meeting Minutes Generator, Project: ASR Shootout - Indian Conversational Speech
- XGBoost: self-declared, 3 evidence items, linked to Project: Fraud Detection MLOps Platform, Project: Health Insurance Fraud Detection System, Project: Credit Card Fraud Detection System
- Zerodha Kite Connect: self-declared, 1 evidence items, linked to Project: AI-Powered Investment Portfolio Platform
Shipped work
See all workFlagship systems this candidate architected will appear here once attached to the dossier. No project has been added yet.
Getting better, on the record
Proven skills over time — today marked in lime.
This trajectory populates once the candidate completes scored interview sessions. No score has been recorded or estimated here.
Career chronology
Gaps shown honestly — only what's on the record.
Great Learning
PGP
Great learning
PGP
Great Learning
PGP
Trainee – Manufacturing & Formulation
CIPLA Ltd
Trainee – Manufacturing & Formulation
CIPLA Ltd
Data Scientist
CIPLA Ltd
Bangalore Institute of Technology
B.E.
Bangalore institute of technology, Bengaluru
B.E.
Bangalore Institute of Technology
B.E.
Hiring for Data Scientist?
Request the full evidence dossier — verified interview sessions, cryptographically attested artifacts, and reference outcomes.
Goes straight to Shreya. No recruiter spam.