Client-Intake-Agent

• 

  Diagram

• 

Inspiration Manual sales triage kills momentum. When enterprise prospects submit unstructured chat payloads about their requirements, sales engineers and SDRs waste hours cross-referencing intent, budget, and technology stacks. I was inspired to eliminate this bottleneck by building an autonomous, zero-latency microservice that handles scope-qualification instantly, freeing up human capital for actual deal-closing.

What it does Client-Intake-Agent is a headless FastAPI microservice that acts as an automated SDR. It ingests messy, conversational prospect messages via a REST API, deterministically extracts canonical tech stacks (e.g., mapping "fastapi" to "FastAPI"), calculates a budget-based qualification score out of 100, and immediately assigns a Lead Status (like QUALIFIED). It then automatically persists these audit-ready JSON records into an embedded database for future CRM syncing.

How I built it Core API Framework: Built the asynchronous backend using Python, FastAPI, and Uvicorn for high-performance routing.

Engine & Logic: Programmed a custom rules-engine using Regex to parse unstructured text, extract specific tech keywords, and mathematically compute qualification scores based on budget thresholds.

Persistence Layer: Integrated SQLite and SQLAlchemy to commit immutable, time-stamped lead records with deserialized JSON arrays.

Testing: Engineered a 100% passing automated test suite using pytest with isolated temporary database fixtures to ensure zero regressions.

Cloud & Infrastructure: Packaged the application into a multi-layer Docker container, pushed the image to Amazon Elastic Container Registry (ECR), and deployed it live as a serverless task on AWS ECS Fargate.

Challenges I ran into One of the main challenges was ensuring the tech-stack extraction engine remained deterministic and accurate despite the countless ways a client might spell or format their technology requirements. Additionally, configuring the AWS ECS Fargate networking cluster to successfully pull the local Docker images from ECR required careful IAM and security group management.

Accomplishments that I'm proud of I am extremely proud of achieving a 100% green automated test suite and successfully containerizing the application for a true cloud-native AWS Fargate deployment. Moving from local development to a live, production-grade cloud environment seamlessly was a major milestone.

What I learned I deepened my expertise in asynchronous Python backend development, temporary database testing fixtures in pytest, and cloud infrastructure containerization. I also learned how to bridge the gap between software engineering and sales operations by designing an API that directly solves a business-revenue problem.

What's next for Client-Intake-Agent The next phase includes integrating external LLMs (like AWS Bedrock or Claude) for deeper natural language understanding, and building automated Webhook triggers to instantly sync qualified leads directly into CRMs like HubSpot or Salesforce.

Built With

• amazon-web-services
• aws-ecr
• aws-ecs
• aws-fargate
• backend
• docker
• fastapi
• microservices
• pytest
• python
• rest-api
• sqlalchemy
• sqlite
• uvicorn