Sailesh Dwivedy's Personal Website

I build AI systems that have to earn trust, not just accuracy: claims agents an auditor will sign off on, clinical notes a paramedic can rely on mid-shift. As an AI/ML Engineer with 5+ years building and deploying production ML, GenAI, and multi-agent systems across insurance, healthcare, and finance, I focus on making these systems auditable, explainable, and reliable enough for high-stakes, real-world use, not just accurate in a demo.

You can connect with me on LinkedIn

Experience

ML Engineer, Central Mutual Insurance (Mar 2026 – Present)

Building production multi-agent AI systems for insurance claims. Architected a Claims Leakage Auditor that coordinates eight specialist agents across live claim and policy data to cut manual review time by 90%, and a Claims Summarizer that generates source-grounded liability reports, cutting report generation time by 83%. Focused on reliability at scale: human-in-the-loop verification, resilient agent harnesses, and full observability into cost, latency, and failure.

AI/ML Engineer Intern, NeuroSpring (Jun 2025 – Mar 2026)

Built ML and GenAI systems for healthcare, centered on speech analysis and clinical documentation. Developed slurred-speech detection models reaching 85% F1 with SHAP-based interpretability for clinical trust, and fine-tuned a clinical LLM to help paramedics draft field notes and retrieve guidelines in real time.

Volunteer Graduate AI/ML Researcher, Image and Video Computing Lab, University of Colorado Boulder (May 2024 – Dec 2024)

Researched knowledge distillation for CNNs and Vision Transformers, co-authoring a NeurIPS 2025 paper on logit-free training. Designed a novel layer-selection metric that improved accuracy by 15% and cut training time by 80%.

Volunteer NLP Research Associate, Kalinga Institute of Industrial Technology (Feb 2023 – Aug 2023)

Conducted applied research on transformer-based document parsing for business documents, benchmarking modern layout-aware models against OCR and earlier baselines and achieving 5 to 10% accuracy gains in structured field extraction.

Associate Software Engineer II (Data Science), HighRadius Technologies (Jul 2019 – Feb 2023)

Led a 10-member team delivering production ML systems for Fortune 500 clients including Nestlé, Uber, and Kraft Heinz, spanning fraud detection, payment prediction, and cash forecasting. Improved dollar recovery by 12%, reduced DSO by 30%, and mentored 50+ interns.

Tech Skills

Languages & ML: Python, SQL, PyTorch, Scikit-learn

Agentic AI & RAG: LangChain, LangGraph, Deep Agents, MCP, Vector Search, Evaluation & Observability

GenAI & NLP: Hugging Face, Transformers, Fine-tuning (LoRA/QLoRA), RLHF

MLOps & Cloud: AWS, Azure, GCP, Docker, Kubernetes, FastAPI, MLflow, Weights & Biases

Education

M.S. Computer Science, University of Colorado Boulder (GPA 3.90/4), Aug 2023 to May 2025

B.S. Computer Science, KIIT University, May 2016 to May 2020

Happy to connect. Always up for a conversation about applied AI, agentic systems, and building them for the real world. Reach me at sdwivedy30@gmail.com or on LinkedIn.