Dhruv Prajapat
I design and ship AI systems — LLM agents, retrieval pipelines, and multimodal products that hold up in production.
Building retrieval systems that reason, not just fetch — and vision models that run at the edge.
I'm an AI engineer who cares about the boring parts — latency, evals, drift — because that's what makes intelligent products feel effortless.
I ship LLM-powered assistants, agentic workflows, and retrieval systems that move business metrics. I like small teams, sharp problems, and prototypes that survive contact with real data.
- LLM apps & retrieval
- Agentic workflows & tools
- Evals, guardrails & tracing
- Serving & inference infra
Things I've shipped
Atlas RAG
Retrieval-augmented assistant over 2M+ enterprise documents with hybrid search and grounded citations.
Iris Vision
Real-time anomaly detection on manufacturing lines with a fine-tuned ViT + edge inference.
Signal Recs
Two-tower recommender with contextual re-ranking that lifted CTR by 18% on a media catalog.
Nova-7B
Domain-tuned 7B model with LoRA + DPO for financial reasoning, served on vLLM.
Toolkit
Timeline
- 2024 — NowAI Engineer· Independent
Designing LLM agents and retrieval systems for early-stage teams.
- 2022 — 2024AI Engineer· Stealth Startup
Owned retrieval + ranking pipeline serving 50k daily users.
- 2018 — 2022University· B.Tech, Computer Science
Thesis on self-supervised representation learning.
Let's build something intelligent together.
Open to consulting, contract, and full-time roles in AI.