Antara Raman Sahay

Designed and developed an intelligent onboarding agent to improve user activation and feature adoption. Analyzed onboarding funnels, user behavior, and conversational interactions to identify drop-off points. Implemented a ReAct-based agent architecture for user intent understanding and proactive recommendations. Built contextual recommendation workflows and personalized engagement strategies using conversational memory. Architected and deployed a scalable agent platform integrating 1000+ external toolkits. Designed a dependency graph-based tool intelligence layer and built a Planner Executor Evaluator agent architecture using LangGraph. Developed a ReAct-based Tool Selection Agent and implemented backend infrastructure using LangGraph, PostgreSQL, MySQL, and LangSmith. Engineered support for large-scale tool chaining, execution monitoring, and enterprise-grade agent operations.

Optimized multi-agent system for software test case execution, reducing execution time by 40%. Deployed a caching-based replay mechanism using OpenAI SDK and PocketBase. Configured a mode-switching execution strategy and reduced redundant LLM calls. Architected a customer-facing multi-agent QA chatbot suite with adaptive memory, artifact persistence, and multi-modal ingestion. Productionized the system with ARQ-based guardrails, caching, and retries.

Delivered live interactive sessions on core and advanced Data Science topics including Python, SQL, Machine Learning, Deep Learning, and MLOps.

Led development of an AI-powered code review system for pull requests, reducing manual review time. Conducted research on academic and commercial solutions, fine-tuned models, and improved performance. Designed and released a CI-ready Python package and FastAPI model server, integrated with Bitbucket Pipelines and Jenkins. Built an NLP-powered dashboard for drilling data analysis and a memory-based multimodal RAG pipeline. Structured and optimized an OCR pipeline for PDF data extraction using Azure AI and layout-aware models. Benchmarked LLMs and VLMs for accuracy and reliability. Architected a custom Graph RAG pipeline and fine-tuned a Small Language Model for enterprise data retrieval. Engineered a Jenkins Shared Library PoC for ML and LLM pipelines.

Researched and deployed language model architectures to address business challenges using NLP. Processed and visualized large unstructured datasets, extracting actionable insights and improving decision-making efficiency. Delivered scalable AI/ML solutions, reducing processing time and enhancing model accuracy.