Mrunal
Fulzele

.NET & AI Engineering | Freelance Independent Projects

2025 — Present

1. Built an agentic Irish public-transport assistant using .NET, Microsoft Agent Framework, MCP tool-calling and TFI GTFS Realtime feeds, solving the challenge of accessing live routes, delays and next departures through a conversational interface. Implemented input/output token management and AI security guardrails, including prompt-injection defence and least-privilege tool access, enabling reliable real-time journey responses while controlling AI context, cost and tool-access risks.
2. Architected and deployed a .NET 8 microservices backend with Ocelot API Gateway, centralising service routing and aggregation to establish a scalable and maintainable application architecture. Containerised and productionised the platform using Docker on a secure VPS, implementing HTTPS and subdomain-based routing to deliver a publicly accessible, production-ready environment.

Led implementation within an 8-engineer, multi-vendor Agile team, owning low-level design for the customer-journey API and mentoring 2 junior engineers, improving delivery capability through structured code reviews and pair programming.

1. Re-engineered a .NET Core REST API processing 2M+ requests/month, optimising Dapper/SQL Server access paths and introducing Azure Service Bus asynchronous processing to reduce response time from 800ms to 300ms (62%) while maintaining contractual SLAs.
2. Developed .NET Core ingestion APIs processing telemetry from 1,000+ Windows and ChromeOS endpoints, integrating SQL Server and Angular to deliver role-based administration and audit logging for a large-scale monitoring platform.
3. Converted founder and customer requirements into production features, defining release scope and developing responsive Angular components in a startup-style delivery, accelerating alignment between business priorities and technical delivery.
4. Managed Azure DevOps and Git CI/CD pipelines within Scrum and multi-vendor delivery, standardising release processes to support consistent deployments and reliable delivery against agreed SLAs. Translated client requirements and technical trade-offs for non-technical stakeholders to facilitate solution alignment across vendors.

Achievements: Reduced API response time by 62% (800ms → 300ms) for a client-facing .NET Core REST API processing 2M+ requests/month through Dapper/SQL Server optimisation and Azure Service Bus asynchronous processing. Delivered an end-to-end monitoring platform for 1,000+ endpoints, implementing .NET Core ingestion APIs, SQL Server integration and an Angular admin dashboard with role-based access and audit logging.

Served as Senior Engineer on a 6-person product team developing KonnectInsights, a multi-tenant social listening and engagement SaaS platform built on ASP.NET MVC/Web API, SQL Server and Azure, supporting 40+ enterprise brands.

1. Owned the platform’s event-driven ingestion architecture, designing and implementing the Twitter Premium API pipeline using Azure Service Bus topics, subscriptions, dead-lettering and idempotent consumers, enabling near-real-time processing of 500K+ social events per day into the unified inbox.
2. Re-engineered the social-event ingestion pipeline to reduce mention-to-inbox latency from 15-minute batches to under one minute, improving response-time SLAs and reducing dropped and duplicated events to near zero.
3. Implemented tenant-level data isolation and GDPR-aligned social-data handling across the multi-tenant platform, strengthening data security and supporting compliant processing of client and end-user information.
4. Evaluated and presented architecture trade-offs, including topic fan-out versus queues and throughput versus cost, to product and client-success stakeholders, while improving release stability through Azure DevOps CI/CD automation and environment standardisation.

1. Delivered core features for CloudSocial, a multi-tenant social media SaaS integrating five social platforms, supporting publishing and social listening for 100+ agencies as part of a 5-person engineering team.
2. Eliminated manual sentiment triage by architecting a Python NLP microservice integrated with C#, automatically classifying thousands of inbound messages daily and delivering production AI-powered classification in 2021.
3. Engineered REST API integrations across Google, Meta and other social platforms, implementing Swagger contracts, API versioning, OAuth and webhook processing to enable reliable automated publishing and social listening.
4. Improved backend reliability through rate-limit handling, structured error management and retry mechanisms, reducing disruption from third-party API failures and strengthening stability of client-facing workflows.