Question 1

Who is Kaushik Saravanan?

Accepted Answer

Kaushik Saravanan is an AI engineer and MS in Artificial Intelligence Engineering candidate at Carnegie Mellon University (ECE, expected December 2027), based in Pittsburgh, PA. He was previously a Developer Associate at SAP Labs India (2024–2026), where he shipped production RAG and agentic AI systems to 400+ users. IEEE-published researcher and Smart India Hackathon 2022 winner.

Question 2

Is Kaushik Saravanan open to new AI/ML roles?

Accepted Answer

Yes. Kaushik is open to Summer 2027 AI/ML and RAG internships in the US, and full-time AI engineering roles starting January 2028 after his CMU MS-AIE graduation. Reach out via LinkedIn (linkedin.com/in/kaushiksss) or X (@Kaushiks0).

Question 3

What did Kaushik build at SAP Labs India?

Accepted Answer

At SAP Labs India (2024–2026) he built a production RAG pipeline over 2M+ indexed vectors serving 400+ users at under 2s p95 latency, fine-tuned DeBERTa for German PII extraction at 94% recall, co-developed three agentic AI systems for SAP HANA operations with LangGraph over MCP tool servers, rewrote a credential-fetch client in dependency-free Go for 9,000+ Linux servers, and executed a zero-downtime multi-cloud consolidation decommissioning 350+ servers for $88K in one-time savings.

Question 4

What are Kaushik's IEEE publications?

Accepted Answer

Two IEEE papers: 'Swarm Intelligence-Based Cooperative Intelligent Transportation System' (ICCIES 2025) and 'Cognitive Intrusion Detection System in Autonomous Vehicles Using Machine Learning' (ICPECTS 2024).

Question 5

What is Kaushik's tech stack?

Accepted Answer

Python, SQL, Go, TypeScript, PyTorch, Hugging Face (TRL, GRPO, SFT, LoRA), DeBERTa, Qwen3, RAG, vector search, ChromaDB, LangGraph, MCP, FastAPI, React, Google Cloud Platform, and Microsoft Azure. Specializes in RAG pipelines, LLM/post-training fine-tuning, and agentic AI systems.

Tagged EarlyWatch

Reading a 40-page EarlyWatch PDF in one glance

Other topics

Tagged EarlyWatch

All posts

Reading a 40-page EarlyWatch PDF in one glance

Other topics