Sarvesh Patil

About Sarvesh Patil

AI Engineer | Generative AI Innovator | Computer Vision Specialist | Data-Driven Problem Solver I’m Sarvesh Patil, an AI Engineer with 4+ years of experience building and deploying intelligent systems that solve real-world challenges. My expertise spans Generative AI, Computer Vision, Telecom Analytics, and Web Data, with a strong foundation in machine learning, data pipelines, and cloud-based deployment. I specialize in designing end-to-end ML pipelines, orchestrating multi-agent systems, fine-tuning LLMs, and implementing deep learning solutions using Python, TensorFlow, and PyTorch. With hands-on experience across industries such as Automotive, Telecom, and Web Data, I combine technical depth with problem-solving skills to deliver scalable, production-ready AI solutions. Key Achievements: • Engineered multi-agent pipelines with CrewAI & MCP, cutting research-to-insight time by 60% and boosting RAG accuracy by 20% • Delivered multimodal AI agents (text, images, PDFs) with fine-tuning loops, improving downstream accuracy by double digits • Developed a YOLOv8-based Sign Language Detection system, achieving 94% accuracy and enabling real-time gesture-to-text conversion • Built a generative AI-powered medical chatbot with LangChain & GPT-4, deployed via AWS CI/CD, reducing latency by 35% and achieving 92% accuracy • Automated high-volume subscriber analytics at Infosys for 11M+ users, cutting churn and MTTR while improving on-time delivery to 95%+ Skills: Programming Languages: Python, SQL Frameworks: Flask, Streamlit, LangChain, LangGraph, LangFlow, CrewAI, Agno Data Analytics: ETL processes, data validation, data warehousing, predictive analytics, Alteryx Data Visualization: Power BI (DAX, Power Query), Tableau, Excel (Pivot Tables, VLOOKUP) Machine Learning & AI: Regression, Classification, Clustering, Recommendation Systems, Predictive Modeling, Time Series Forecasting, Pattern Recognition, NLP, Hugging Face Transformers, OpenCV, Responsible AI (SHAP, LIME), Model Monitoring Databases: MySQL, PostgreSQL, MongoDB, Apache Hive, Neo4j MLOps & Workflow Automation: MLflow, Airflow, Git, Docker, REST APIs, Model Deployment, CI/CD, A/B Testing Cloud Platforms: Microsoft Azure (Data Factory, Synapse, Databricks, Azure ML), AWS (S3, EC2, SageMaker) Tools & Project Management: Jira, Git, Agile, ServiceNow, Postman, VS Code 🤝 Let’s Connect: Email: s••••••••@gmail.com

Experience

1. Ai Engineer

   Bright Data

   May 2024 — Present · New York, NY, US

   Built multi-agent pipelines with Crew AI and MCP servers to orchestrate LLM agents for scraping, RAG, and context handover, cutting research-to-insight time by 60% and boosting answer freshness in production. Designed agentic retrieval (chunking, vector search, context windows) with session-aware handovers and grounding, improving RAG accuracy by 20% and reducing hallucinations in long-running workflows. Built robust data ops: cleaning and standardizing data, loading agents that fill vector stores and OLTP/OLAP databases, and scheduling “freshness timers,” enabling near‑real‑time updates (minutes, not hours) and over 99% pipeline uptime. Delivered multimodal agents for text, images, and PDFs, plus fine-tuning and pre-training loops on curated datasets, improving downstream task accuracy by double digits and expanding coverage to previously unstructured content. Engineered self-repairing agent workflows using Crew AI and custom orchestration logic, with retry chains, anomaly detection, and fallback strategies—cutting failure rates by over 85% and keeping multi-agent pipelines resilient under volatile conditions. Integrated live observability with Prometheus, Grafana, and internal MCP monitoring to track agent health, latency, and data drift—making it easy to catch issues early and consistently across global deployments.

Education

• Pace University - Seidenberg School of Computer Science and Information Systems

  Master of Science - MS

  2023 — 2025