Testimonials
Services
Video meeting . 15 mins
Discovery Call for Students
Free 15 min call for students starting out in AI
Video meeting . 15 mins
5
Video meeting . 30 mins
5
AI Engineer Interview Prep
Interview prep on RAG, agents and LLM evaluation
About me
Yash Kumar | AI Engineer and Mentor
I build AI systems that real people use every day. At PayNearby I work on Saathi AI, a voice financial counselor on Google ADK and Gemini Live that serves 1,500+ users.
What I've built
Saathi AI: six domain sub agents covering financial products and app support. I raised its RAG accuracy from 70% to 90% by rewriting the knowledge base around a schema, measured by manual review of 1,000+ responses.
Evaluation pipeline: self hosted Opik with a custom tracer for ADK's streaming path, a 109 question golden dataset, and per domain LLM as judge scoring.
Fine tuning: LoRA adapters on Gemini 2.5 Flash through Vertex AI, one per domain.
GeniOS: founding AI engineer on a LangGraph pipeline that resolved 92% of queries across 500+ concurrent users.
CricVision: a four agent cricket analytics pipeline, built in 3 hours at GDG Cloud New Delhi.
Recognition
1st prize, Qubit Quest 2025, IIIT Delhi. Top 13 of 4,000+ teams, HackChrono 2025. Speaker at the Rox Community Event, Bengaluru, on context engineering.
What you get in a session
Gen AI roadmap: what to learn, in what order, and which projects get noticed.
Project review: feedback on your RAG or agent project from someone who has shipped one to real users.
AI engineer interview prep: what interviewers ask about RAG, agents and evaluation.
How I mentor
Every piece of advice comes from something I shipped or something that broke in production. You leave with one concrete next step for your own project.
Let's connect
Book a session or message me on LinkedIn.