Bibhanshu Raj
@bibhanshuraj
AI engineer building intelligent systems — from voice agents and RAG pipelines to production ML infrastructure. Turning research into products.
Building AI That Ships
Not just research — production systems that scale.
I'm Bibhanshu Raj, an AI engineer currently at Makunai Global. I specialize in building end-to-end machine learning systems — from training custom models to deploying them at scale.
My work spans voice agents, RAG pipelines, and production ML infrastructure. I believe the best AI isn't the most complex — it's the one that reliably ships value to users.
When I'm not building voice agent systems or optimizing inference pipelines, you'll find me exploring new ML research and contributing to practical AI engineering.
What I've Built
A selection of AI-powered products and tools.
Production Voice Agent Platform
Real-time STT/TTS voice agent with Deepgram + OpenAI response generation. RAG pipeline with Pinecone for context-aware voice conversations. FastAPI + WebSockets backend with sub-second response times. Handles concurrent sessions at scale.
AI-Powered Evaluation Bot
95% grading accuracy across 200+ student submissions. Automated 85% of evaluation steps, cutting grading time by 60%. Intelligent rubric matching and automated feedback generation.
Multi-Document RAG Chatbot
Contextual Q&A over 500+ academic documents with 95% retrieval accuracy. 40% improvement in user satisfaction through claim validation. Vector similarity search for efficient document retrieval.
LLM-Powered Learning Agent
RAG architecture for precise answers from research papers. NVIDIA NeMo Guardrails for content safety and quality. Custom RAGAS metrics for response quality evaluation.
Tech Stack
The tools and technologies I work with daily.
AI / Machine Learning
Languages
Where I've Worked
From building voice agents to scaling AI at production.
Makunai Global
AI Engineer & Lead Developer
Architecting production-grade voice agent systems — end-to-end pipeline from speech recognition to response generation. Building scalable backend powering real-time voice AI using FastAPI, WebSockets, and cloud-native services.
SmartLink Holdings
Engineering Intern
Engineered a warehouse inventory system using Python and PostgreSQL, reducing parts logging time by ~40%. Validated a computer vision-based PCB error detection tool using live camera feeds.
Education
Academic background and coursework.
BITS Pilani
B.Sc. Computer Science & Information Technology
Let's Build Something
Have a project in mind? I'm always open to discussing new opportunities.