AI Engineer at Merck Group · Building Intelligent Systems
I turn raw data and ideas into AI systems that actually work ; systems that learn, reason, and deliver real results. From building multi-agent AI assistants to designing pipelines that save millions in costs.
Python
TensorFlow
Docker
FastAPI
AWS
PostgreSQL
Node.js
Neo4j
Grafana
Git
Scikit-learn
PyTorch
GitHub
TypeScript
Next.js
Firebase
n8n
LangflowPythonTensorFlowDockerFastAPIAWSPostgreSQLNode.jsNeo4jGrafanaScikit-learnPyTorchGitHubTypeScriptNext.jsFirebasen8nLangflow
I'm an AI Engineer with over 3 years of hands-on experience designing and shipping intelligent systems ; from conversational agents to large-scale data pipelines. Currently at Merck Group, where my work directly saves millions in operational costs.
My focus is on making AI practical and impactful: systems that answer real business questions, automate tedious manual work, and give teams superpowers they didn't have before. I've led projects from concept to production across supply chain, brand intelligence, document processing, and more.
I hold a B.E. in Computer Science, certifications from Microsoft, Coursera, and LinkedIn Learning, and a published research paper on computer vision for visually impaired users.
Real projects, real outcomes. Here's what I've built and the numbers behind it.
A selection of projects that pushed the boundaries of what AI can do in the real world.
Led the design of an AI recommendation system that uses semantic understanding and graph databases to match products to people ; boosting product visibility by 70%. Deployed via CI/CD with Docker in a fully automated pipeline.
Built a retrieval-augmented generation system that reads documents (PDFs, Word files, plain text) and generates accurate product descriptions using GPT-4. Makes unstructured data actionable and reduces manual writing effort significantly.
Implemented a high-speed semantic search system using the Weaviate vector database ; making it 70% faster to find relevant information. Combined AI vector search with traditional keyword search to improve accuracy by over 40%.
Built a real-time object detection and navigation system using computer vision and reinforcement learning to help visually impaired users move safely through complex environments. Published as a research paper.
Built a scalable pipeline that ingests, processes, and makes sense of large volumes of live data ; improving throughput by 40%. Integrated with machine learning models to surface actionable business insights continuously.
A personal project that turns "happy birthday" into something that actually means it. Relationships are modeled as a graph in Neo4j, a living web of who-loves-who, because that is genuinely the right shape for that kind of tenderness. OpenAI helps write each wish, and Firebase verifies that the phone number typing it in really belongs to the person sending it.
Visit wishyoubirthday.comThe languages, frameworks, and platforms I use to build intelligent systems.
Continuous learning from the world's top platforms.
Peer-reviewed research contributing to accessible AI technology.
Trained YOLOv7 on the COCO dataset (330k images, 80 object categories) to build a real-time object detection system that helps visually impaired users understand and navigate their environment safely. The research evaluates the model on accuracy, speed, and memory footprint ; demonstrating that cutting-edge AI can be made accessible and practical for people who need it most.
A portfolio shouldn't only be read, it should be played too. Jump the runtime errors before they crash your run.
Click or tap to play. Space or the up arrow also works.
Whether you have a project in mind, want to discuss AI solutions, or just want to say hello, I'd love to hear from you.
Claude Code
Codex
Replit Agent
OpenCode