Hey there, I'm
Yash Dabhade — I build AI systems.
I like building AI that people actually rely on at work every day. Right now that's at Reno, where I own the AI vertical end to end. One of my favourite builds is our ops agent: the team just chats with it to get live details from the database, schedule tasks, and monitor operations, without digging through dashboards.
Most of my work sits where models meet messy real-world systems: retrieval that holds up in production, agents that don't fall over on the tenth step, and evals that tell you whether a change actually helped. I've been hands-on from day one, from backend work at an early-stage startup to being a founding engineer, and I still enjoy owning a problem from the first prototype to the on-call fix.
These days, voice AI is where most of my curiosity goes. Alongside work, I'm pursuing an M.Tech at BITS Pilani, still love a good hackathon, and write here about what I'm building, reading, and learning.
ElsewhereGitHub ↗LinkedIn ↗LeetCode ↗
~/now
- Building
- Audio 90, Code-Eagle AI
Latest writing
All writing →- Building Voice Agents: Cascaded Pipelines, Speech-To-Speech, and the War on Latency
A voice agent is a chatbot where every 100 ms is visible. How the cascaded STT → LLM → TTS architecture works, how it compares to native speech-to-speech models, and the streaming and turn-taking tricks that make it feel human.
- LightRAG: Graph RAG Without the Graph-Sized Bill
GraphRAG is powerful but slow and expensive to build and update. LightRAG keeps the knowledge graph, drops the community summaries, and retrieves at two levels — specific entities and broad themes — at a fraction of the cost.
- GraphRAG: When Your Answers Live Between the Documents
Vector search finds passages that look like your question. It cannot answer "what are the main themes across all of these?". GraphRAG builds a knowledge graph and community summaries so it can.
Building
All projects →Audio 90
Pre-market testing for audio ads: compare two creatives and see a directional winner, the reason, and the uncertainty before buying media.
Code-Eagle AI
End-to-end pull-request review system using Qwen-Coder, QLoRA, Tree-sitter AST context, and static analysis.
Achievements
- Winner, xAI D.O.G.E International Hackathon (StateShift)
- LeetCode Knight · 1942 max rating (top 3%)
- 1st runner-up, Smart India Hackathon 2022 Grand Finale
Education
- M.Tech, Artificial Intelligence & Machine Learning
- B.Tech, Computer Engineering · 9.30 CGPA
Interests
Audio & voice AI agentsRetrieval & evaluationAgent reliabilityMultimodal systemsLLM infrastructureCompetitive programming