Siva Rami Reddy Kandula

- Tech lead for Finance AI inside ServiceNow's Office of the CFO. Build forecasting, agentic-AI, fine-tuning, and AI-governance tooling adopted across financial planning and operational reporting at a Fortune 500 SaaS company. Lead 3 engineers. - Forecasting — Shipped a neural forecasting engine using Temporal Fusion Transformers (TFT) and DeepAR on multivariate financial time-series; reduced forecast MAPE by ~15% vs. the ARIMA / ETS baseline, with quantile outputs for probabilistic planning. Hierarchical reconciliation across global / regional forecasts; macroeconomic features as exogenous regressors. - Agentic AI & RAG — Built agentic workflows on Llama 3/4 + LangChain for autonomous anomaly detection on financial transactions and spend patterns; cut manual triage by ~40%. Designed RAG with hybrid dense + sparse retrieval (BM25 + embedding rerank) and grounded-citation enforcement; +30% factual-response accuracy on internal evals. - LLM Fine-Tuning — Applied LoRA / QLoRA via the PEFT library for domain-specific finance tasks where base + RAG was insufficient; tuned for retrieval-aware generation and structured-output reliability. - MLOps & Observability — Production stack on Kubeflow + MLflow with CI/CD, model registry, online + offline serving, retraining triggered by drift signals. LLM observability on Langfuse (tracing + eval) + Grafana + Prometheus for system / model metrics; sustained 99.9% inference availability. - AI Governance — Operationalized ServiceNow AI Control Tower for Finance AI — model inventory, drift / concept-shift monitoring, model cards, eval harnesses. Engineered the model-risk path for LLM features: bias / fairness checks, grounding evals, and audit-ready documentation aligned to NIST AI RMF and EU AI Act risk-tiering principles. - Strategy — Authored a build-vs-buy analysis for the forecasting stack (in-house TFT/DeepAR vs. AWS Forecast / DataRobot); contributed to roadmap and feature definition for Now Assist for Finance.

Senior member of the AI Search team building retrieval infrastructure for MyADP, a regulated HR / payroll platform serving ~1B annual logins. - Engineered Lucene-based search indexes powering the AI Search backend; optimized query performance for Dense Passage Retrieval (DPR) models serving millions of concurrent users. - Built a precision-evaluation framework using SBERT sentence embeddings; full-coverage offline evaluation that quantified retrieval quality and drove product roadmap decisions. - Contributed to backend scalability of the MyADP application during peak payroll cycles — sharding, caching, and query-path optimization to maintain high availability under load.

Machine Learning engineer specializing in computer vision and automated pipelines for geospatial data analytics. Key Deliverables & Tech Stack:Computer Vision: Co-developed a Deep Learning Convolutional Neural Network (CNN) for automated geospatial site mapping, achieving 96% benchmark accuracy. DevOps / MLOps: Introduced production CI/CD pipelines and modular architectures, accelerating three major product release cycles by 2 months each. Recognition: Awarded Star Employee of the Year 2020 for excellence in predictive analytics delivery.

Trainee- AI/ML