Usman Babar

AI Engineer specializing in scalable backend systems and LLM integration.

Based in: Islamabad, PakistanCurrently: AI Engineer, Howie FlexCoExperience: 1 year

Hire me

Usman is a Data Science student and AI Engineer with extensive experience building production-grade REST APIs and data ingestion pipelines using FastAPI and Node.js. He has a proven track record of implementing RAG systems and orchestrating complex workflows with Temporal and vector databases like Qdrant.

Experience

AI Engineer

Howie FlexCoVienna, Austria (Remote)contractJun 2025 – Present

Summary

AI Engineer - Full-Time Contractor

What I did

  • Developed and maintained backend REST API services using FastAPI (Python), integrating PostgreSQL, Elasticsearch, and Qdrant to support a production data ingestion pipeline
  • Built and shipped an ML classification service end-to-end, including API design, database integration, and workflow orchestration using Temporal
  • Debugged and resolved a recurring backend pipeline timeout by implementing an asynchronous Temporal activity, improving pipeline reliability
  • Implemented server-side data validation logic prior to database writes across PostgreSQL, Elasticsearch, and Qdrant, improving data integrity
  • Implemented a workflow for correct file selection and data extraction.
  • Fine-tuned YOLOv11x for title block detection in drawings used in the data pipeline.

Results

  • Trained an ML classifier on 50,000 files and 40 GB of data to distinguish between drawing and non-drawing firm data.
  • Redesigned the ingestion pipeline into an event-based architecture where classification, processing, and data sanitization are isolated and repeatable.
FastAPIPythonPostgreSQLElasticsearchQdrantTemporalREST APIML classificationYOLOv11x

Data Analyst & Developer Intern

Civic IQ (USA Startup)Remote / HybridinternshipJul 2025 – Aug 2025

Summary

Data Analyst & Developer Intern

What I did

  • Built backend automation scripts in Python and Node.js to ingest and process data on Google Cloud Platform (GCP)
  • Integrated third-party large language model APIs to automate contract data analysis and validation, reducing manual verification effort
  • Engineered a data validation pipeline for extraction workflows, ensuring high data integrity across automated jobs
  • Automated data validation and ingestion for a civilian data service business.

Results

  • Eliminated approximately 70% of manual verification work by implementing an NLP solution to detect organization entities.
  • Automated a data extraction process that previously required 10 data analysts working for 7 days.
  • Processed 100 GB of term extraction data in 5 days.
PythonNode.jsGoogle Cloud Platform (GCP)LLM APIsData ValidationNLP

Skills

Technical

FastAPIFastAPI
PythonPython
LLM APIs
Qdrant
VS Code
Third-party API integration
REST APIs
GitGit
DockerDocker
Jupyter Notebooks
JavaScript (Node.js)JavaScript (Node.js)
SQLSQL
Express.jsExpress.js
Node.jsNode.js
PostgreSQLPostgreSQL
Elasticsearch
Web Scraping
Data Validation
ETL
Debugging & Testing
Playwright
RedisRedis
Cheerio
Google Cloud Platform (GCP)Google Cloud Platform (GCP)
BeautifulSoup
MongoDBMongoDB
C++C++
RR
Event-based architecture
ML Classification
NLP
TypeScriptTypeScript
Zod
LoRA
Fine-tuning
YOLOv11x
TypeScript GenericsTypeScript Generics

Projects

Type-safe integration wrapper

personal
Built a wrapper to prevent frontend crashes caused by silent API payload changes from the backend.
TypeScriptZod

Internal AI Knowledge Base

Backend Developerpersonal
Built a self-hosted platform with FastAPI exposing OpenAI-compatible REST API endpoints, containerized with Docker for deployment. Integrated third-party AI model APIs (Ollama, LlamaIndex) and used Redis for caching to improve response performance.
PythonFastAPIDockerRedisQdrantOllamaLlamaIndex

LearnX - AI-Powered Video Learning Platform

Full Stack Developerpersonal
Built a full-stack web application with a Node.js/Express backend exposing REST APIs, MongoDB for data storage, and a React frontend. Developed web scraping and content ingestion pipelines using Cheerio and Playwright to automate collection of source material for the platform. Integrated a context-aware RAG chatbot API endpoint using LlamaIndex and Qdrant to deliver time-synced answers grounded in video transcripts.
Node.jsExpressMongoDBReactCheerioPlaywrightLlamaIndexQdrantRAGLoRAPEFTQwen

Education

Bachelor of Science

FAST-NUCESData ScienceGPA 3Aug 2022 – Jun 2026
Hire UsmanReach out about a role, a contract or a conversation.For recruiters

Usman's twin is AI, it can make mistakes.

More people on Kanvis

Browse more peopleCreate your own Kanvis