Fine-tuned transformer-based LLMs (BERT, GPT-4) on over 500,000+ domain-specific financial documents (e.g., earnings

reports, SEC filings), boosting sentiment classification accuracy and entity recognition F1-score by 32% through iterative

hyperparameter tuning and stratified cross-validation.

 Spearheaded LSTM model deployment for multi-class sentiment prediction across social media and financial news feeds,

improving F1-score by 35% over rule-based baselines using attention layers and advanced NLP pipelines.

 Leveraged AWS Lambda with API Gateway to deploy scalable & secure GenAI endpoints, accelerating automated reporting

& personalized client communications by 40%, while ensuring end-to-end compliance with FINRA and SEC mandates.

 Optimized training and inference processes via AWS Nova, reducing compute costs by 45% through GPU instance tuning, load

balancing, and auto scaling policies without affecting system latency or model accuracy.

 Developed and automated CI/CD pipelines using GitHub Actions and Docker, cutting deployment time, enhancing rollback

reliability, and ensuring 99.9% production uptime for financial AI services.

 Designed ML-based automation for retirement plan approval, reducing processing time for low-risk applicants by 20%, while

ensuring full compliance with fiduciary standards through interpretable risk scoring models.

 Streamlined NLP workflows using LangChain, automating the parsing and summarization of financial documents with a 70%

reduction in manual review time, improving consistency in client reporting and templated communications.

Designed and implemented deep learning models using PyTorch and TensorFlow to forecast IT infrastructure failures,

reducing critical system downtime by 35% through proactive anomaly detection and real-time alerting systems.

 Developed automated document classification models with CNNs and attention-based mechanisms, streamlining over 10,000+

monthly IT service tickets and reducing manual triage effort by 60%.

 Built and deployed real-time ML pipelines on Azure using Blob Storage and Azure ML for cybersecurity threat detection,

increasing threat identification accuracy by 40% and enabling daily model refresh cycles.

 Created scalable ETL pipelines that processed over 7 million rows of IT operational logs, significantly improving data quality

and consistency across analytics and model training workflows.

 Optimized Azure ML workflows for automated model training, validation, and hyperparameter tuning, cutting model

development time and improving IT asset forecasting precision by 28%.

 Developed interactive Power BI dashboards to track model accuracy, visualize IT performance KPIs, and monitor security

anomalies, resulting in 20% faster decision-making for leadership and audit teams.

 Engineered a multi-modal recommendation system for IT resource allocation using transformer architectures and

collaborative filtering, improving resource utilization efficiency by 25% and reducing provisioning delays by 30%