Siddhesh Inamdar

Lead Data Scientist | M.Tech, IIT Madras | BITS Pilani

I believe the most powerful data-driven solutions aren’t just calculated—they’re crafted. To me, the real magic happens at the bleeding edge where algorithmic precision meets high-performance engineering. I dedicate my time to deconstructing frameworks, optimizing low-level kernels, and bridging the critical gap between complex AI research and scalable, production-ready systems.

Expertise & Domains

• Generative AI & Agentic Workflows: Specialized in architecting production-grade Model Context Protocol (MCP) agents leveraging tools like CrewAI and LangChain. I frequently implement Mixture of Experts (MoE) routing patterns to dynamically optimize and specialize model responses.
• Kernel & Inference Optimization: Deeply experienced in squeezing out maximum performance using custom Triton GEMM kernels, Post-Training Quantization (PTQ), and TensorRT-LLM to achieve massive inference speedups for large-scale models.
• Computer Vision: Proficient in fine-tuning state-of-the-art vision architectures, including adapting YOLOv11 on complex aerial imagery to accurately identify and predict dynamic fleet cluster movements.
• Optimization & Probabilistic Modeling: Expert in formulating robust discrete programming optimizers (CP-SAT) for massive-scale fleet scheduling, alongside building highly-calibrated MCMC-based models for rigorous sensor data reconciliation.

Professional Journey

Lead Data Scientist | Caterpillar

Jul 2025 – Present

Currently spearheading AI initiatives for the Marine Division, with a hyper-focus on leveraging computer vision to revolutionize maritime supply chain visibility and orchestrating complex agentic workflows. Beyond core model development, I architect and manage our team’s cloud infrastructure. Recently, I engineered highly-scalable retrieval pipelines that seamlessly integrate Milvus Vector Databases with Snowflake, empowering operators with real-time insights for strategic data-driven decisions.

Data Scientist | ExxonMobil BTC

Aug 2021 – Jun 2025

Architected and deployed high-impact, enterprise-scale AI solutions. Key highlights include formulating a sophisticated CP-SAT optimizer to coordinate operations for over 900 autonomous vehicles, and building a robust RAG-based conversational interface powered by Llama 3 to drastically streamline maintenance log querying. My tenure also involved engineering Temporal Fusion Transformers (TFT) to deliver advanced, highly accurate oil production forecasting.

Technical Toolkit

Let’s Connect

I am always eager to discuss the frontiers of artificial intelligence, the evolution of autonomous agents, and the intricacies of high-performance data architecture.

Feel free to connect and reach out via my social handles!