GenAI Engineer - Risk & Compliance Automation

Thapaswi Movva

LinkedIn 318-***-**** ***************@*****.***

WORK EXPERIENCE

Navikenz Houston, USA

AI/ML engineer Nov. 2024 - present

● Designed and deployed a production-grade GenAI platform for credit risk and compliance teams, enabling retrieval-augmented insights across loan policies, KYC documents, transaction histories, and regulatory guidelines to support analyst decision-making.

● Built secure data ingestion and preprocessing pipelines for structured risk data and unstructured financial documents, applying chunking, metadata enrichment, and access controls to ensure compliant and role-based retrieval in regulated environments.

● Implemented RAG workflows combining vector-based semantic retrieval with prompt orchestration and relevance filtering, delivering grounded, evidence-backed responses while minimizing hallucinations in risk and compliance use cases.

● Developed AI-assisted investigation and case summarization workflows, enabling analysts to quickly review historical cases, transaction patterns, and policy interpretations while maintaining human-in-the-loop approvals for final decisions.

● Integrated AI-generated insights directly into CRM and case management systems, enriching customer, loan, and investigation records with retrieved policy context, transaction summaries, and risk indicators to streamline analyst workflows.

● Built secure API integrations between the GenAI platform and CRM systems, enabling AI-assisted case summarization, evidence retrieval, and recommendation previews within existing business workflows while preserving audit trails.

● Engineered low-latency AI inference services and APIs supporting search, summarization, and Q&A use cases, optimizing response times and cost efficiency under concurrent internal usage.

● Established LLMOps practices, including prompt versioning, retrieval evaluation, response-quality monitoring, and failure analysis, to ensure reliability, reproducibility, and auditability of GenAI systems in production.

● Deployed AI services using Docker and Kubernetes, implementing autoscaling, retries, and fallback strategies to meet operational SLAs in regulated FinTech environments.

● Implemented monitoring and audit logging to track retrieval sources, response usage, latency, and system errors, supporting compliance reviews and internal audits.

● Authored technical and compliance documentation detailing system architecture, data flows, CRM integrations, access controls, and AI usage boundaries for risk governance teams. Tiger Analytics Hyderabad, IN

ML engineer July 2022 - Apr. 2023

● Designed and maintained scalable data engineering pipelines on Azure Databricks (PySpark) to ingest, cleanse, and transform large-scale image, text, and interaction datasets, handling schema evolution, partitioning, and data quality checks to support downstream ML workloads.

● Built end-to-end ML pipelines to generate ResNet50 image embeddings and structured text features, storing intermediate representations in optimized analytical formats and constructing ANN indexes for similarity search, improving precision@10 by 15% on imbalanced datasets.

● Developed and productionized learning-to-rank and recommendation systems, including SVM, XGBoost, and DSSM models for search relevance and ALS, RNN, and LSTM models for session-based recommendations, delivering ~10% CTR uplift and 25% reduction in false negatives versus baseline approaches.

● Engineered feature preprocessing and aggregation pipelines that unified offline training data and online inference inputs, ensuring feature consistency across batch and real-time workflows and improving model stability in production.

● Designed and deployed production-grade ML inference services exposing /search, /recommend, and /similar APIs, containerizing models with Docker and orchestrating deployments on AKS, sustaining 1,000+ QPS with p95 latency under 200 ms.

● Implemented MLOps workflows using MLflow and Azure Machine Learning, enabling experiment tracking, model versioning, registry-driven deployments, and scheduled retraining with drift detection to maintain performance as data distributions evolved.

● Built reproducible CI/CD pipelines in Azure DevOps, automating data pipeline validation, model packaging, and environment-specific deployments while enforcing rollback and versioning standards.

● Integrated ranking and recommendation outputs into customer-facing applications in collaboration with product and frontend teams, executing controlled A/B experiments to validate relevance improvements and business KPIs.

● Added runtime monitoring and logging for data pipelines and ML services, tracking data freshness, processing latency, service throughput, and prediction distributions to support reliability and rapid debugging in production.

● Authored technical documentation covering data ingestion workflows, feature pipelines, model architectures, API contracts, and deployment procedures, enabling maintainability and smoother team onboarding. KFinTech Hyderabad, IN

ML engineer June 2020 - June 2022

● Led the design and automation of an end-to-end risk analytics data platform, building 12+ batch and near–real-time pipelines using Azure Data Factory and PySpark to integrate KYC, loan, and transaction datasets into Azure Data Lake and Azure Synapse, reducing reporting turnaround time by 25% and enabling compliance teams to access refreshed data within 10 minutes.

● Developed and operationalized credit risk prediction services using Logistic Regression, Random Forest, and XGBoost on Azure Machine Learning Studio, improving default prediction accuracy from 70% to 80% and reducing manual underwriting effort by 20% through automated decision support.

● Architected a real-time fraud detection pipeline on Azure Databricks, applying Isolation Forest, One-Class SVM, and Autoencoders to streaming payment data, detecting anomalous transactions in <200 ms and reducing false positives by 10%, strengthening fraud monitoring SLAs.

● Built customer segmentation and behavioral modeling pipelines using K-Means and DBSCAN across 250K+ customers, uncovering repayment and transaction patterns that informed risk-based loan pricing and targeted repayment strategies.

● Containerized ML inference services using Docker and deployed them on Azure Kubernetes Service (AKS), enabling horizontal autoscaling up to 2,000 requests per second while maintaining p95 inference latency under 180 ms for downstream risk and fraud applications.

● Designed a custom feature engineering framework (offline + online parity) for domain-specific signals such as transaction velocity, repayment-to-income ratio, and merchant risk, minimizing training–serving skew and improving model stability in production.

● Established CI/CD pipelines with Azure DevOps and Terraform, standardizing infrastructure, data pipelines, and ML deployments across environments and reducing production rollbacks by 20%.

● Implemented model and pipeline observability using Prometheus and Grafana, tracking data freshness, pipeline health, prediction distributions, and accuracy drift, enabling proactive intervention and consistent SLA adherence across risk analytics systems.

● Authored technical documentation and runbooks for data pipelines, ML workflows, and production deployments, enabling smoother onboarding and operational handoffs. PROJECTS

Cloud-based Basketball Ranking Data Tracker Ruston, LA Cloud Composer Bigquery Cloud Function Looker Cloud Storage Nov. 2024

● Automated the fetching of data daily once from the rapid API tool with cloud-composer and then DAG was implemented with a function to trigger the dataflow job to pipeline the data from the bucket .

● The cloud function is triggered upon the file uploaded to the bucket to trigger the dataflow job. The dataflow job has the option to load the data from the storage bucket to the BigQuery. Airline Ticket Management System Ruston, LA

MySQL SQL Shell Database Optimization Aug. 2024

● Designed a MySQL database to manage airline tickets, flights, passengers, and staff details.

● Used SQL to retrieve available seats, calculate flight revenue, and manage bookings and cancellations.

● Automated ticket booking, seat allocation, and payment updates in the database. Crime Against Women Prediction System Ruston, LA

Python Pandas scikit-learn Matplotlib Data Augmentation Regression Models Nov. 2024

● Created a dataset combining government records with data augmentation techniques to handle missing and imbalanced data by SOMTE

● Added socio-economic features such as literacy, income to improve prediction quality.

● Trained and tested Logistic Regression, Random Forest, Linear Regression and achieved about 80% accuracy in identifying high-risk regions.

SKILLS & CERTIFICATIONS

Cloud & Databases: Azure (Data Factory, Synapse, Databricks, ML Studio, AKS, Data Lake Gen2) Snowflake PostgreSQL MongoDB Vector Databases (FAISS, Pinecone) Data Engineering & Orchestration: PySpark Apache Airflow Azure Functions Azure Logic Apps Terraform Azure DevOps MLflow

Machine Learning & AI: scikit-learn XGBoost SVM Random Forest CNN (ResNet) RNN/LSTM TensorFlow PyTorch keras

Large Language Models & GenAI: GPT-4 Retrieval-Augmented Generation (RAG) Prompt Engineering LangChain LLM Fine-tuning

MLOps & Model Lifecycle: MLflow Azure ML Pipelines Model Registry CI/CD Drift Detection Monitoring

(Prometheus/Grafana)

EDUCATION

Louisiana Tech University Ruston, LA

Master of Science in Computer Science March 2025

● Coursework: Software Engineering, Adv. Analysis of Algorithms, Adv. Computer Architecture, Cloud Computing, Data Mining, Computer Networks, Project Management Jawaharlal Nehru Technological University Hyderabad, IN Bachelor of Technology in Electronics and Communication Engineering May 2022

● Coursework: Database Management,Machine Learning, Digital Image Processing, Statistics, Data Communication and Networks.

Contact this candidate