[email protected] · Mumbai, India · aryanrajpurkar.com · LinkedIn · GitHub

B.Tech CSE (Data Science) · D.J. Sanghvi College of Engineering · CGPA 9.34 / 10 · 2022–2026

I design and ship production-grade AI systems — agentic pipelines, large-scale data platforms, and hybrid search infrastructure. My work spans:

• Autonomous AI agents — multi-tool, RAG-backed, LLMOps-ready
• Data engineering — real-time & batch ETL, streaming (Kafka/Airflow), 100K+ daily records
• Search & retrieval — BM25 + vector hybrid, knowledge graphs, re-ranking

Currently: AI Engineer Intern @ Atlan · Freelance AI Engineer @ Aretis Labs · Founding Engineer, Data Platforms @ VisaFriendly

Highlights

• Scaled AI automation consultancy to 10+ enterprise clients across industries in 5 months (Aretis Labs)
• Built ETL pipelines processing 100K+ monthly job entries with 28% accuracy improvement (VisaFriendly)
• Designed multi-agent evaluation pipeline automating 8K documents/month, improving interview response rates 22%
• Architected private on-premise RAG knowledge engines improving brand citation rates 30% across LLMs
• Engineered LangGraph recruitment workflows with 40% latency reduction and 20% lower cost per candidate

Intelligent document intelligence platform with precedent relationship graph modeling provenance, dependencies, and citation lineage across government documents. Enables graph-traversal queries over document metadata.

• Hybrid RAG: BM25 + FAISS vector retrieval, cross-encoder re-ranking, compliance checks via Apache Kafka
• MCP-published catalog operations (hybrid search, contradiction analysis, version chains, grounded Q&A)
• Stack: Next.js · Python (Flask) · MongoDB (GridFS) · Neo4j · BM25 · FAISS · Elasticsearch · OCR · Kafka

Agentic AI content studio with DAG-based orchestration across 6 CrewAI LLM agents. Citation-style provenance ties every output to specific sources with retrieval quality gates.

• Multi-model image generation (Flux, Nano Banana) as agent-callable tools
• RAG pipeline with BM25 sparse + dense vector retrieval, configurable top-k and 0–100 relevance scoring
• Stack: TypeScript · React · CrewAI · ChromaDB · FastAPI · DALL-E · Stable Diffusion · WebSockets

Enterprise AI evaluation and safety layer with multi-layer detection (toxicity, bias, hallucination, jailbreak) backed by a fine-tuned toxicBERT running 500+ adversarial test cases.

• SHAP/LIME explainability via REST API for enterprise audit trails
• Multi-tenant architecture with per-tenant guardrail configs
• Stack: Python (FastAPI) · React/TypeScript · MongoDB · PyTorch · Transformers · SHAP · LIME · Redis · Docker

Languages

AI / ML

Web & APIs

Cloud & Data

DevOps

• Winner — Smart India Hackathon 2024 · National hackathon by the Government of India
• Amazon ML Summer School Scholar 2024 · Selected from 61,000 applicants nationwide
• Best Student Chair 2025 · Society of Data Science, India — organised 10+ workshops
• Chairperson, DJS-S4DS · National Data Science Committee of India, mentored 200+ students
• 5× National Hackathon Winner · ₹2L+ prize pool

Open to high-impact roles in AI Engineering, Data Engineering, and Agentic Systems.
[email protected]