Greater Ahmedabad Area
Bad RAG hallucinates. Bad ad AI wastes client budgets. I've built systems that do neither. I'm an AI Engineer focused on LLM pipelines, RAG architectures, and agentic workflows that ship to production, not just demos. M.Tech in Data Science, PDEU - completed June 2026. At Schbang Digital Solutions, I built a multimodal generative AI platform processing 500+ ad creatives with 95%+ structured output accuracy and generating ad assets with 98% Meta compliance prediction, protecting real client spend. For my thesis, I built a GraphRAG conversational AI system over 30 research papers - hybrid retrieval with embeddings + BM25, BGE cross-encoder reranking, knowledge graph traversal, and RAGAS-evaluated at 0.85+ faithfulness. Both shipped. Both measured. Both documented. Both in Production. Stack: LangChain · FastAPI · Qdrant · Pinecone · FAISS · Gemini · Claude API · Redis · Docker. Open to AI Engineer, ML Engineer, or Applied NLP roles. [email protected]
Architected a multimodal generative AI platform (Gemini Pro Vision + LangChain + RAG) extracting 50+ features across 500+ ad creatives, delivering automated brand audits and creative recommendations, reducing brand research turnaround from ~2 weeks to under 4 hours. Built a Meta ad compliance prediction engine achieving 98% approval accuracy - eliminating an estimated 40-50 rejected submissions per month and saving in wasted client ad spend. Engineered production agentic LangChain workflows with 3-tier retry logic, prompt engineering and Pydantic validation, automating brand strategy workflows from consumer intelligence extraction to compliant creative generation that previously required 35-60 hours of manual analyst work per campaign. Developed a consumer segmentation and ad-to-audience alignment system across 35+ behavioral dimensions, reducing creative testing cycles by 40% and improving campaign ROAS through precision targeting. Shipped an AI-driven creative generation pipeline producing production-ready ad assets deployed in live client campaigns, and contributed to the frontend of getadvize.ai using React, Next.js, and Tailwind CSS.
Graded assignments for 120 students across DBMS, Operating Systems, and Cloud Computing courses under faculty supervision, maintaining consistent evaluation standards across all three subjects. Conducted faculty-guided research applying graph-based AI to traffic management systems and computer vision with object detection to road network classification
Built and evaluated supervised ML models including Logistic Regression and Random Forest, and applied unsupervised techniques for pattern discovery, using cross validation and hyperparameter tuning to optimize performance across structured datasets. Executed end to end data preprocessing pipelines covering EDA, feature engineering, and data cleaning to ensure model ready inputs across multiple projects. Implemented NLP workflows including text preprocessing, TF-IDF vectorization, and classification, and communicated findings through structured visualizations and summary reports.