Applied AI Engineer Gen AI & Production ML Systems

Deepak Rambarki

Applied AI Engineer Generative AI & Production ML Systems

Denton, TX +1-940-***-**** **************@*****.*** LinkedIn GitHub Portfolio SUMMARY

Applied AI Engineer with 5+ years of experience building and scaling production ML and LLM systems serving 3M+ daily transactions. Delivered measurable business impact including 19% fraud reduction and 28% revenue uplift through real-time inference, advanced modeling, and cloud-native MLOps on AWS. Specialized in Retrieval-Augmented Generation (RAG), agentic workflows, low-latency inference architecture, and responsible AI governance across the full ML lifecycle. Impact Snapshot: 19% fraud reduction • 28% revenue uplift • 3M+ daily transactions • 15TB+ data scale • Sub-100ms inference • 35% retrieval improvement

SKILLS

Programming: Python (Advanced), SQL, R, Java, C++, OOP, Data Structures & Algorithms Machine Learning: Classification, Regression, Anomaly Detection, Time Series Forecasting, XGBoost, LightGBM, TensorFlow, PyTorch, CNNs, LSTM/BiLSTM, Transformers, Transfer Learning, Hyperparameter Optimization (Optuna, Bayesian), QLoRA, PEFT Generative AI & NLP: Large Language Models (LLMs), GPT-4/5, LLaMA 2/3/4, Retrieval-Augmented Generation (RAG), LangChain, LangGraph, Prompt Engineering, Embeddings, Semantic Search, ChromaDB, BLEU/ROUGE, Hugging Face, spaCy Data & Big Data: Apache Spark (PySpark), Kafka, Airflow, ETL Pipelines, Feature Engineering, Snowflake, Redshift, MongoDB, Cassandra MLOps & Deployment: AWS SageMaker, GCP Vertex AI, Azure ML, MLflow, CI/CD for ML, Model Monitoring, Data Drift Detection, FastAPI, Docker, Kubernetes, Real-Time Inference APIs

Cloud Platforms: AWS, Azure, GCP, AWS Lambda

Evaluation & Responsible AI: AUC-ROC, F1-Score, Precision@K, SHAP, Model Interpretability, AI Governance, GDPR/HIPAA Awareness EXPERIENCE

AI Engineer, Brex Dec 2024 – Present

Designed and deployed a real-time fraud detection system processing 3M+ daily transactions, reducing fraud by 19% using BiLSTM- based behavioral sequence modeling and ensemble methods.

Architected end-to-end ML pipelines from data ingestion to production inference using AWS SageMaker, Snowflake, and microservices architecture.

Built scalable feature engineering pipelines with Python, SQL, and PySpark, processing 10M+ records and managing 15TB+ structured data.

Improved model performance to 0.87 F1-score / 0.93 AUC-ROC through hyperparameter optimization and ensemble modeling.

Built agentic RAG workflows using LangChain and LangGraph, orchestrating multi-step LLM reasoning pipelines with retrieval grounding (ChromaDB), structured outputs, and evaluation guardrails to support automated fraud case triaging.

Implemented QLoRA-based fine-tuning and PEFT strategies on LLaMA models to optimize domain adaptation while reducing training costs.

Developed FastAPI-based inference services and containerized deployments (Docker, Kubernetes), achieving scalable sub-100ms latency.

Automated retraining, model versioning, and monitoring using MLflow, CI/CD pipelines, Airflow orchestration, and drift detection.

Implemented SHAP-based explainability and AI governance controls, supporting audit readiness and regulatory compliance. Machine Learning Engineer, Zebronics India Jun 2020 –May 2023

Developed a hybrid recommendation engine (collaborative + content-based filtering) increasing cross-sell revenue by 28% across 3M+ users.

Designed personalized ranking algorithms leveraging behavioral embeddings, improving recommendation accuracy by 20% over baseline.

Built demand forecasting models using XGBoost, LSTM, ARIMA, and Prophet, reducing overstock by 18% and stockouts by 22%.

Designed ML-powered REST APIs integrated into web and mobile platforms, improving conversion rates by 10% and session duration by 12%.

Conducted structured A/B testing and experiment design, driving 6% retention growth and measurable revenue uplift.

Developed real-time ML pipelines using Azure ML, Kafka, Spark Streaming, Docker, and Kubernetes, enabling scalable automated deployment.

Optimized inference performance and resource utilization, reducing infrastructure costs while maintaining high availability.

Partnered with product and business stakeholders to translate KPIs into measurable ML objectives aligned with revenue strategy.

Built data quality monitoring and drift detection pipelines to proactively prevent production model degradation. PROJECTS

Financial Advisor LLM github.com/deepak-rambarki/financial_advisor_llm Built an LLM-based conversational advisory system using structured prompting and contextual grounding to generate personalized portfolio insights and compliance-aligned outputs.

Implemented retrieval grounding, response evaluation pipelines, and guardrails to reduce hallucinations and ensure consistent financial recommendations.

MedCompare — AI Medication Benchmarking Tool

Developed an LLM evaluation platform benchmarking medication outputs using semantic similarity, fuzzy matching, BLEU/ROUGE metrics, and FHIR-compatible exports.

Designed automated evaluation dashboards enabling systematic comparison of multiple models and improving reliability for healthcare AI use cases.

Model Interpretability Framework github.com/deepak-rambarki/model-interpretability-for-machine-learning-models Designed an audit-ready explainability toolkit using SHAP and LIME to generate feature attribution reports and production model diagnostics. Built monitoring dashboards highlighting feature drift, prediction confidence, and bias indicators to support regulatory readiness. XGBoost Classifier + ONNX Edge Deployment

Developed a large-scale predictive model (1.2M records) outperforming baseline classifiers and optimized preprocessing for high-dimensional data.

Converted the model to ONNX and optimized inference latency for scalable edge and cloud deployment in IoT environments. EDUCATION

Master in Advanced Data Analytics Aug 2023 – May 2025 University of North Texas Denton, TX (GPA - 3.8/4) Relevant Coursework: Deep Learning, Recurrent Neural Networks, Natural Language Processing, Big Data Systems, Advanced Analytics Deployment

Contact this candidate