Mr. Mitul Ashvinbhai Trivedi

Anomaly Detection in Automobile Supply Chain Management: Leveraging iForest & Autoencoders for Risk Mitigation

Abstract:

Supply chain disruptions cost the global automotive industry an estimated $210 billion in 2021, with 84% of companies reporting parts shortages or delayed shipments. Traditional monitoring methods fail to detect hidden irregularities such as fraudulent invoices, shipment delays, and warehouse stock imbalances until financial and operational damage occurs. This session explores how Random Isolation Forest (iForest) and Autoencoder-based deep learning models can proactively detect anomalies across procurement, logistics, and inventory. iForest identifies rare, high-risk events like 5–10% cost spikes in supplier contracts or outlier shipment delays exceeding 3σ thresholds through recursive partitioning in high-dimensional datasets. Meanwhile, Autoencoders reconstruct baseline patterns from millions of IoT sensor readings or demand forecasts, flagging deviations when reconstruction errors surpass defined thresholds. Case studies reveal that integrating these models reduces manual monitoring efforts by up to 40%, enables real-time fraud detection in <2 seconds per transaction, and cuts stockout risks by 30–50%. Moreover, combining anomaly detection with predictive analytics enhances supplier reliability scores by 25%, leading to improved on-time delivery performance. Attendees will gain a practical framework for embedding anomaly detection into existing supply chain ecosystems, using real-world metrics to demonstrate ROI in fraud prevention, cost optimization, and service-level improvements. In an era of global volatility, this approach empowers automotive firms to create data-driven, resilient, and adaptive supply chains. 

Profile:

Mitul Trivedi is a seasoned Data Scientist with over 15 years of experience in Generative AI, Machine Learning, and Deep Learning, delivering impactful AI solutions across diverse industries. Currently leading Generative AI initiatives in the legal domain at Tata Consultancy Services Limited, he specializes in developing multilingual AI products for document automation, legal research, and summarization.

Mitul's technical expertise spans RAG architectures, LLM fine-tuning, prompt engineering, and hallucination detection, with hands-on experience using LangChain, Pinecone, and AWS Bedrock. Prior to this, he served as Lead Data Scientist at Google, where he applied AI-driven legal analytics and predictive modeling to support the Global Affairs team using GCP Vertex AI.

His earlier leadership at APM Terminals focused on AI/ML strategy, data governance, and financial analytics, driving operational efficiencies across global port operations. A Certified Databricks Professional, Mitul holds an MBA in Technology (Data Science) from Walsh College and advanced credentials from The University of Texas at Austin and Case Western Reserve University.

Proficient in Python, SQL, Spark, and major cloud platforms (AWS, Azure, GCP), Mitul combines deep technical knowledge with strategic business insight to deliver high-impact, scalable AI solutions.