Rajani M.

Senior AI Engineer

LTIMindtree

Aug 2024 - Current (2 years 1 month)

Designed an AI system to detect fraudulent transactions in real-time payment systems. Collected & labeled 10M+ transaction data points. Implemented deep learning anomaly detection model with 95% precision. Integrated solution into AWS Lambda for real-time monitoring. Impact: Reduced fraudulent transaction losses by 40% within 6 months. Responsibilities: Designed and implemented data ingestion pipelines for 10M+ real-time transactions using SQL + Python ETL. Applied feature engineering (transaction velocity, user geolocation, device fingerprinting). Trained deep learning model (Autoencoder & CNN in TensorFlow/Keras) for anomaly detection. Deployed fraud detection service as a real-time API on AWS Lambda & S3. Integrated alerts with busines

Autonomous Litigation Research & Brief Drafting Agent

LTIMindtree

Feb 2024 - Current (2 years 7 months)

• Developed a multi-agent LegalAI system for law firm clients, enhancing efficiency in case law research. • Reduced junior associate research time from 3-4 hours to under 45 seconds per task. • Designed a 5-node LangGraph state machine architecture on Azure AI Foundry for intelligent task routing. Key results: • 10× faster case law precedent discovery vs. manual Westlaw research • Bad-law citation rate reduced 8% → 0.3% via real-time LexisNexis Shepardization • 94% attorney satisfaction in 200-query blind evaluation • Zero jailbreak or data-leak incidents across 90-day production window

C

Autonomous Litigation Research & Brief Drafting Agent

Cyril Amarchand Mangaldas CAM

Jul 2023 - Dec 2023 (6 months)

• Developed a multi-agent LegalAI system for law firm clients, enhancing case law research efficiency. • Reduced junior associate research time from 3-4 hours to under 45 seconds per task. • Utilized a 5-node LangGraph state machine architecture on Azure AI Foundry for intelligent task routing.

W

Generative AI Engineer

Watermark Insights

Jun 2022 - Jul 2024 (2 years 2 months)

Built NLP model to analyze and classify customer reviews from e-commerce platforms. Developed sentiment classifier (Positive/Neutral/Negative) using BERT. Processed 1M+ reviews and visualized insights using Power BI. Exposed REST API for integration with customer support systems. Impact: Improved customer experience analysis, boosting CSAT score by 20%. Responsibilities: Collected and preprocessed 1M+ e-commerce reviews using NLTK, spaCy (tokenization, stopword removal, lemmatization). Trained BERT-based NLP classifier with Hugging Face Transformers for sentiment categories. Built visualization dashboards in Power BI to showcase customer sentiment trends. Developed and deployed Flask REST API to integrate predictions with customer service

W

Principal

Watermark Insights

Jun 2022 - Jul 2024 (2 years 2 months)

Built NLP model to analyze and classify customer reviews from e-commerce platforms. Developed sentiment classifier (Positive/Neutral/Negative) using BERT. Processed 1M+ reviews and visualized insights using Power BI. Exposed REST API for integration with customer support systems. Impact: Improved customer experience analysis, boosting CSAT score by 20%. Responsibilities: Collected and preprocessed 1M+ e-commerce reviews using NLTK, spaCy (tokenization, stopword removal, lemmatization). Trained BERT-based NLP classifier with Hugging Face Transformers for sentiment categories. Built visualization dashboards in Power BI to showcase customer sentiment trends. Developed and deployed Flask REST API to integrate predictions with customer service

AI-Powered Financial Knowledge Assistant (RAG Platform)

PwC

Jun 2022 - May 2023 (1 year)

• Developed an enterprise RAG platform for context-aware Q&A over financial documents and compliance policies. • Automated document ingestion pipeline processing over 50,000 PDFs and SEC filings with zero manual intervention. • Enhanced semantic search precision by 31% using a hybrid BM25 and dense retrieval approach with Cohere Rerank. Key results: • Hallucination rate reduced to <4% on financial policy Q&A (measured via Ragas faithfulness) • 50,000+ documents ingested at launch — fully automated • Human-in-the-loop escalation UI flags low-confidence responses for compliance review • Deployed on Azure with FastAPI backend, Docker containerisation, and MLflow tracking Azure OpenAIAzure AI SearchRAGLangChainPineconeCohere Rerank

Machine Learning Engineer

SitusAMC

May 2021 - Jul 2022 (1 year 3 months)

Designed an end-to-end automated ML pipeline from data ingestion to deployment Implemented data and model versioning using DVC and Git Built REST API using FastAPI for real-time predictions Integrated MLflow for experiment tracking and model registry Developed CI/CD pipelines using GitHub Actions for automated testing and deployment Containerized application using Docker for reproducibility Deployed model on cloud platform with scalable infrastructure Set up monitoring for data drift and model performance degradation Enabled automated retraining pipeline based on drift detection

Team Lead

SitusAMC

May 2021 - Jul 2022 (1 year 3 months)

Developed predictive ML models using IoT sensor data from manufacturing equipment. Built ETL pipelines in PySpark to handle streaming IoT data. Trained LSTM model to forecast failures with 92% recall. Automated retraining workflow using Airflow and Azure ML pipelines. Impact: Reduced downtime by 25% and saved $1.5M in maintenance cost

Multimodal AML Anomaly Detection & MLOps Pipeline

SitusAMC

May 2021 - Jun 2022 (1 year 2 months)

At SitusAMC, I engineered a sophisticated multimodal ML system to identify illicit financial transactions by integrating diverse data signals. This innovative approach replaced traditional compliance engines, resulting in a significant reduction in false positives and improved detection accuracy. My work included implementing a full MLOps pipeline, ensuring robust monitoring and retraining capabilities, which led to a six-month production run without undetected data drift incidents.

Test Lead

HCL Technologies

Jun 2020 - May 2021 (1 year)

Built time-series forecasting model for retail sales to support demand planning

Full Stack Engineer

HCLTech

Jun 2020 - May 2021 (1 year)

Built time-series forecasting model for retail sales to support demand planning

Real-Time Fraud Detection Platform

HCL Technologies

May 2020 - Apr 2021 (1 year)

Designed and deployed a real-time fraud classification platform combining ML ensemble models with SHAP explainability and an LLM-powered semantic investigation interface. Ran structured experimentation across Logistic Regression, Random Forest, XGBoost, and Gradient Boosting — selecting XGBoost by precision-recall tradeoff analysis. Built a synthetic test data generation framework producing 50K fraud scenarios across 8 attack patterns for shadow-mode evaluation before live rollout. Key results: • F1 = 0.91 on held-out fraud test set • 100% of model decisions auditable via SHAP feature-importance reports • Shadow-mode validated against 6 months of historical production logs pre-launch • <2s response latency on LLM-powered investigator Q&A i

Senior Test Engineer

Cigniti Technologies

Dec 2019 - Sep 2020 (10 months)

2+ Years’ experience in UI Development and backend Development using languages Java, JavaScript, Junit, TestNG, JIRA. WhiteBox testing using Java, JavaScript based framework. Understand System design and Low level design components

Truffaut CELUM - DAM Phase 1 and Phase 2

Cognizant

Dec 2017 - May 2018 (6 months)

Truffaut DAM is used to Asset management as used maintain the asset. CELUM - The CELUM Digital Asset Management helps to improve processes.

Associate - Project (Platform JnJ)

Cognizant

Dec 2016 - Dec 2017 (1 year 1 month)

Description: Smart Trial application is used to manage and processing Clinical data as well as used to find Risk. Responsibilities: • Establish an automated test environment for UI and non-UI testing • Preparation of automation Test scripts using Selenium Web driver. • Design, document, manage and execute test cases, sets, and suite • Review test cases and automate whenever possible • Idea in Test Case Prioritization • Better knowledge in Designing the Test Data Properly • Debugging and running the Tests

Senior Associate

Cognizant

Dec 2014 - Sep 2019 (4 years 10 months)

Built ML models to predict customer churn for a telecom company using historical usage and billing data.

SmartTrials 2.0 Roche integration

Cognizant

Dec 2014 - Mar 2015 (4 months)

SmartTrials 2.0 Roche integration is the Roche specific release. Roche data is loaded in to landing layer and transformed in to UI. Responsibilities: • Understanding of the system requirements, low level design and high level design. • Data base testing with SQL queries and API Design and development. • Responsible for Design and testing the API and UI components, white box testing.

Programming Analyst

Cognizant

Aug 2013 - May 2015 (1 year 10 months)

Worked on UI Development HTML, CSS, JavaScript Worked on API Design and Development and WhiteBox Testing

NC Novo Glow

Cognizant

Aug 2013 - Nov 2014 (1 year 4 months)

NC Novo Glow is the ETQ Reliance applications relevant for novoGloW NC. NN aims at reducing the time spend on novoGloW NC. Purpose is to migrate highly paper based system by the system A2DBweb and ODS (Online Deviation system). A2DBweb is web-based and the purpose of the system is to register and control NC and FI, Audit, IC and Inspection reporting, including follow-up reports. Responsibilities: • Understanding of the requirements architecture of the project. System design and implementation Low level design understanding, high level design document understanding. • Responsible for system testing, white box testing, Automation and API testing and regression testing • Develop and Test API and request and response.

About

Detail

Timeline

Résumé