Rohan Menon

Engineered custom Virtual Try-On workflows using open-source models (Flux 2 Klein 4B, Multimodal Gemma E4B) on serverless RunPod GPUs, cutting costs by 90% vs. GPT Image 2.0 to under 1 INR/image. Co-developed full-stack architecture (Python, Node.js, Cursor) and deployed a local MCP server integrating Claude and Gemini to automate bulk ATS CV parsing.

AI Intern - Gen AI Virtual Try on Tool at Wama Technology (2026-07 – Present)

• Engineered custom Virtual Try-On workflows using open-source models (Flux 2 Klein 4B, Multimodal Gemma E4B) on serverless RunPod GPUs, cutting costs by 90% vs. GPT Image 2.0 to under 1 INR/image.
• Co-developed full-stack architecture (Python, Node.js, Cursor) and deployed a local MCP server integrating Claude and Gemini to automate bulk ATS CV parsing.

AI/Python Intern - Instagram AI Analytics Platform at FOREO (2026-02 – 2026-06)

• Built an end-to-end Brand Media Monitoring & Sentiment Intelligence Pipeline to weekly scrape media content from 109 websites, extracting 1,826 skincare-related articles using Scrapy and Playwright.
• Designed ETL workflows to clean structured article metadata and content in PostgreSQL, enabling scalable downstream analytics, and retrieval tasks.
• Developed a RAG-based retrieval pipeline by chunking article data, generating embeddings, and ingesting them into Qdrant Vector DB for efficient semantic search and contextual retrieval.
• Implemented hybrid keyword + semantic search to improve brand-specific document relevance and support accurate sentiment-driven report generation from large-scale media data.
• Built FastAPI backend services integrating Google Gemma (Ollama Cloud), Gemini 1.5 Flash API, to generate automated sentiment reports, and LLM-powered business recommendations
• Developed ML pipelines using Scikit-Learn (Random Forest, Logistic Regression) for engagement prediction and comment sentiment analysis, and built a Streamlit dashboard for Brand sentiment report visualization, analytics monitoring

Machine learning Intern at Font AI (2025-06 – 2025-08)

• Engineered PyTorch pipelines for CLIP encoders and U-Net architectures, executing large-scale model training jobs on AWS SageMaker GPU instances.
• Optimized model throughput and experiment tracking by integrating Flash Attention for improved GPU utilization and Weights & Biases (W&B) for hyperparameter logging and evaluation.

M.Sc. in Neuroscience (Computational track) – University of Bonn (2022-10 – 2025-09)

B.Sc. in Chemistry – University of Mumbai (2019-04 – 2022-04)