Kaif Shaheem

Own the primary real-time search feature powering live product discovery for Pepagora's B2B marketplace, supporting full-text search, semantic retrieval, autocomplete, and typo-tolerant query understanding across 200K+ product records. [OpenSearch, Python] Architected a multi-factor ranking framework combining semantic relevance, geographic proximity, supplier subscription signals, and query intent understanding to deliver context-aware search results. [OpenSearch, Python] Reduced end-to-end search latency by 78% (18.4s → 2.8s) through retrieval pipeline redesign, OpenSearch index optimization, Redis caching, asynchronous prefetching, and parallelized query execution. [Python, OpenSearch]

AI Research Intern at Pepagora - The B2B Growth Engine (2026-03 – Present)

• Own the primary real-time search feature powering live product discovery for Pepagora's B2B marketplace, supporting full-text search, semantic retrieval, autocomplete, and typo-tolerant query understanding across 200K+ product records. [OpenSearch, Python]
• Architected a multi-factor ranking framework combining semantic relevance, geographic proximity, supplier subscription signals, and query intent understanding to deliver context-aware search results. [OpenSearch, Python]
• Reduced end-to-end search latency by 78% (18.4s → 2.8s) through retrieval pipeline redesign, OpenSearch index optimization, Redis caching, asynchronous prefetching, and parallelized query execution. [Python, OpenSearch]
• Implemented hybrid lexical and vector retrieval pipelines to improve product discoverability and ranking quality across diverse supplier catalogs. [OpenSearch, Vector Search, Python]
• Partnered with engineering and product stakeholders on architectural trade-offs to deliver production-ready search capabilities. [Git, Agile]
• Designed and maintained a nightly delta-sync ETL pipeline from MongoDB to OpenSearch, implementing schema normalization and incremental indexing across 200K+ B2B product documents. [Python, MongoDB]
• Built an SVM - based taxonomy mapping model to classify supplier products into standardized categories, improving catalog quality and search discoverability. [Python, Scikit-Learn, SVM]

Backend Software Engineer Intern at MedGlance (Open Source) (2025-05 – 2025-07)

• Architected and deployed cloud-native backend services for authentication, conversation management, and healthcare workflows in a production application. [FastAPI, Firebase Auth, Firestore, GCP Cloud Run]
• Built a multimodal retrieval pipeline combining semantic search and structured medical knowledge to support real-time healthcare question answering. [FastAPI, Qdrant, Gemini]
• Designed RESTful APIs and conversational state-management workflows for context-aware interactions and consistent responses. [FastAPI, Firestore]
• Built AI-powered clinical assistance workflows with contextual memory and retrieval-augmented reasoning. [LangGraph, Gemini, Qdrant]

AI Intern at Katomaran Technologies (2024-05 – 2024-11)

• Architected a multi-tenant multimodal RAG platform for industrial troubleshooting, enabling technicians to query image- and table-heavy equipment manuals while ensuring strict tenant isolation across multiple client organizations. [FastAPI, PostgreSQL, Redis]
• Researched, evaluated, and designed the retrieval architecture, comparing embedding models and multimodal retrieval approaches to identify the most effective solution for visually dense technical manuals. [ColPali, BGE]
• Architected a scalable multi-vector hybrid retrieval and asynchronous document ingestion pipeline, combining semantic and lexical search with queue-based parsing, chunking, embedding generation, and vector indexing for image- and table-heavy industrial manuals. [ColPali, ColQwen2, Python, FastAPI]
• Designed backend strategies for document lifecycle management, including ingestion workflows, chunking strategies, embedding generation, metadata modeling, caching, and retrieval optimization for large-scale industrial knowledge bases. [FastAPI, PostgreSQL, Redis]
• Designed role-based platforms for super administrators, company administrators, and field technicians, supporting manual management and citation-grounded AI troubleshooting assistants. [FastAPI, RAG, Vision LLM]

Quantitative Developer at Right Skew Finalytics (2023-05 – 2023-06)

• Developed quantitative analysis workflows to evaluate financial market trends, patterns, and performance indicators. [Python, Pandas, NumPy]
• Applied statistical modeling techniques to support trading strategy assessment and data-driven investment decisions. [Python, Statistics]

Integrated M.Sc in Data Science – PSG College of Technology (2021 – 2026)