Open to opportunities

BIBHANSHU User

@bibhanshuraj19

AI engineer leading real-time voice AI systems, building scalable RAG and workflow automation with low-latency pipelines.

India

Message

What I'm looking for

I’m looking for a role where I can build low-latency AI systems and production RAG/tool pipelines—owning architecture, performance, and reliability end to end. I want a team that values fast iteration, measurable outcomes, and deep technical ownership.

I build production AI systems that feel fast, reliable, and grounded in real data. At Makunai Global, I scaled a real-time voice AI platform end-to-end (VAD → STT → GPT-4o → TTS) with sub-800ms latency across 100+ concurrent sessions.

I focus on architecture that stays robust under load. I designed a decoupled control/data plane with FastAPI orchestration and horizontally scalable worker processes to improve fault isolation and enable independent scaling of inference workloads.

I also love shipping AI that actually helps users in workflows. I implemented OAuth 2.0 Google Calendar tooling and a multi-tenant RAG pipeline using pgvector (tenant-isolated retrieval) to improve grounding and reduce hallucinations, and I contributed to MakunFlow’s DAG-based automation engine using Redis Pub/Sub and Celery for async execution.

Experience

Work history, roles, and key accomplishments

Current

AI Engineer & Lead Developer

Current

Makunai Global

Dec 2025 - Present (5 months)

Built and scaled a production real-time voice AI pipeline (VAD→STT→GPT-4o→TTS) with LiveKit and Deepgram WebSocket, achieving sub-800ms end-to-end latency across 100+ concurrent sessions. Designed scalable control/data plane orchestration with FastAPI workers and implemented a multi-tenant tool and RAG layer (OAuth 2.0 + pgvector) plus an in-product DAG workflow engine using Redis Pub/Sub and Cele

LiveKit Deepgram fastAPI WebSockets OAuth 2.0 Pgvector Multi Tenant RAG Celery DAG Workflow Orchestration

Software Engineer Intern

SmartLink Holdings

May 2024 - Jul 2024 (2 months)

Developed backend inventory services in Python and PostgreSQL to support 50K+ daily API requests with high availability. Improved performance and reliability by reducing query execution time 60%, increasing throughput 35%, and cutting average API response time 40% using indexing, query restructuring, Redis caching, and asynchronous handling with connection pooling.

Python PostgreSQL SQL Query Optimization Indexing Redis Connection Pooling Performance Tuning

Education

Degrees, certifications, and relevant coursework

BITS Pilani

Bachelor of Computer Science and Information Technology, Computer Science and Information Technology

Pursuing a Bachelor of Computer Science and Information Technology at BITS Pilani, with relevant coursework in areas such as AI, operating systems, and data structures.