Mansi Gusain

Projects

trivago Copilot

April 2024 - February 2026

MLOps Engineering – Deployment Engineering & CI/CD Pipelines for AI/ML Systems

1. Led the deployment of the open-source LibreChat platform as “trivago Copilot” on Microsoft Azure, enabling 700 employees to access enterprise-grade, agentic AI solutions for daily operations, automation, and generative AI tasks.
2. Contributed to trivago’s deployment and lifecycle management of LibreChat on Azure, leveraging Terraform for Infrastructure as Code (IaC) across environments to provision and manage Azure resources (including Azure Container Apps and supporting services) in a repeatable, version-controlled manner aligned with MLOps best practices.
3. Containerized LibreChat and supporting AI services using Docker, managed and optimized Docker images for performance and security, and deployed them to Azure Container Apps to provide scalable, secure, and cost-efficient access to multi-provider LLMs (OpenAI, Azure OpenAI, Anthropic, and other compatible providers).
4. Integrated Azure Cognitive Search to enable search-service, enterprise-aware AI capabilities, and configured secure connectivity, role-based access, and environment-specific settings for compliant, production-ready AI workloads.
5. Implemented CI/CD-style deployment workflows using Git-based version control, automated build and image publishing for Docker containers, and environment promotion strategies for fully automated releases.

Application & Lifecycle Management – AI System Operations

1. Owned end-to-end lifecycle management of trivago Copilot, including promotion of containers and configurations across dev/stage/prod, managing version upgrades, and maintaining backward compatibility to minimize user disruption.
2. Designed and implemented advanced agent actions and tools connecting LibreChat with enterprise systems (Confluence, Slack, Git), enabling workflow automation and context-aware AI assistance directly within existing business tools.
3. Integrated and standardized multiple MCP servers (Git, Slack, Atlassian, Playwright, DeepL, etc.) with consistent configuration patterns, expanding cross-tool automation, retrieval, and orchestration capabilities for agentic AI solutions and handoff between tools and services.
4. Collaborated with cross-functional teams to monitor, troubleshoot, and optimize AI services using Azure logging and basic monitoring (Azure Monitor / Application Insights-style practices), improving performance, reliability, and cost efficiency.
5. Performed ongoing maintenance by rapidly applying new LibreChat releases, patches, and feature updates from the open-source project, handling configuration changes, regression checks, and stability improvements to keep the platform secure and up to date.
6. Contributed to open-source RAG enhancements for Azure Assistants, and refined datasets / fine-tuning strategies based on job- or use-case-specific requirements (JD-aligned), ensuring AI outputs are relevant, accurate, and aligned with business and compliance expectations.

AI Features, Providers & Agentic Workflows – LibreChat & Enterprise Integration

1. Leveraged LibreChat’s unified interface to orchestrate multiple AI providers (OpenAI, Azure OpenAI, Anthropic, and other compatible backends), and designed the deployment to remain extensible to additional providers (including platforms like Foundry) for future multi-LLM routing and failover strategies.
2. Configured and managed AI model selection, system prompts, and safety / policy guardrails to align responses with enterprise guidelines, enabling secure, auditable AI usage across teams.
3. Utilized LibreChat’s advanced agents to support file handling, code interpretation, and API-based actions; implemented custom actions that trigger downstream workflows (e.g., calling internal APIs, updating tickets, querying knowledge bases) to create end-to-end agentic flows.
4. Contributed to agentic handoff patterns where AI agents coordinate between tools (via MCP servers, APIs, and integrations) to complete multi-step business tasks, while exposing clear control points for human review and override.
5. Designed RAG-style workflows combining Azure Cognitive Search and LibreChat’s search / retrieval capabilities to ground LLM responses in internal documentation and knowledge, reducing hallucinations and improving answer relevance for trivago’s use cases.

Projects

MindMitr Mobile App: Powered by GPT-4o, WhatsApp based AI mental health coach

November 2023 - August 2024

AI Infrastructure & Deployment Engineer – MindMitr Mental Health Assistant

1. Designed, developed, and deployed an AI-powered WhatsApp chatbot (“MindMitr”) using OpenAI GPT models to deliver real-time, conversational mental health coaching and support to 50+ active users in a production environment.
2. Built the backend with FastAPI, exposing secure RESTful endpoints for message handling, user management, and webhook processing from Twilio; deployed on Render with environment-specific configurations for secrets, scaling, and observability.
3. Integrated Twilio’s WhatsApp Business API to manage inbound/outbound messaging, webhooks, and session handling, ensuring low-latency, high-availability communication over WhatsApp.
4. Implemented a scalable data layer using Firebase (Firestore/Realtime DB) for user profiles, session state, and chat history persistence, enabling consistent context across sessions and supporting analytics on engagement and conversation patterns.

Application & Lifecycle Management – AI System Operations & User Journey

1. Designed and maintained an intelligent response engine leveraging OpenAI GPT APIs and structured system prompts to generate context-aware, empathetic responses tailored to sensitive mental health queries.
2. Defined and iterated on safety guardrails, including tone guidelines, refusal patterns for high-risk topics, and escalation recommendations to professional help, ensuring responsible AI behavior aligned with mental health use cases.
3. Implemented motivational and scheduled messaging workflows that analyze past conversations and user preferences to trigger follow-up nudges, check-ins, or encouragement at appropriate intervals, increasing retention and long-term engagement.
4. Managed end-to-end user journeys—from onboarding and consent capture through ongoing AI-assisted coaching—ensuring continuity of context via stored histories in Firebase and providing transparent communication about AI limitations and boundaries.
5. Monitored system performance and reliability using platform-level metrics (Render and Firebase monitoring/logs), troubleshooting issues in message delivery, latency, and model responses, and applying iterative improvements to prompts, configurations, and error handling.
6. Curated and refined a self-created dataset of conversation examples and guidance patterns to systematically improve prompt design and model behavior (GPT‑4o), enhancing consistency, safety, and usefulness of responses without over-claiming clinical capabilities.

Projects

Workato

June 2023 - October 2023

1. Custom Connector Development: Designed and developed specialized Ruby-based SDK connectors, enabling secure and seamless integration between WorkWall and the Workato automation platform.
2. Workato Integration & Certification: Engineered end-to-end integration solutions for WorkWall using Workato, including earning Workato Level-3 certification.
3. Salesforce Integration: Built comprehensive integrations between WorkWall and Salesforce via Workato, automating real-time data synchronization of job postings and opportunities.
4. Workflow Automation: Authored advanced Workato recipes and automated critical business processes such as job posting, detail retrieval, and work-object synchronization, significantly increasing operational efficiency.
5. API Engineering: Developed and integrated RESTful APIs to support reliable, real-time communication between WorkWall, Salesforce, and other external applications.
6. Data Mapping & Transformation: Implemented intricate data mapping, trigger conditions, and conditional logic—including advanced formulas and lookup tables—for accurate and efficient data transformation across systems.
7. Scalable System Architecture: Developed and maintained a robust, modular integration architecture that is scalable and supports future enhancements and integrations.
8. Enhanced CRM & Marketplace Connectivity: Improved overall connectivity and data consistency between WorkWall and Salesforce, optimizing customer relationship management and marketplace operations.

Bachelor of Technology- BTech

National Institute of Technology Srinagar Uttarakhand - August 2018 - May 2022

ICSE (10th) / ISC (12th) Certification

Convent of Jesus and Mary , Dehradun - April 2003 - May 2017

Basic to advanced: retreival-augmented generation (rag)

Ai engineer complete course - 2025

Mcp- model context protocol udemy certification

Ai agents with crewai

Workato automation –i

Workato automation- ii-

Workato advanced automation- iii