Pradyumna Yadav
AI Engineer
AI Engineer specializing in agentic systems, AI infrastructure, and production-scale deployments. Currently at FuturePath AI building multi-modal AI products for Fortune 500 clients - from autonomous agents to real-time voice and knowledge systems.
Available for full-time roles and consulting projects
What I Work On
Explore my work
Agentic Systems
RAG & Knowledge Systems
Voice AI
GPU Inference
LLM Infrastructure
DevOps & Infra
Experience
Where I've worked
FuturePath AI
Nov 2024 – PresentAI Engineer
Building real-time voice assistants for enterprise IT operations - phone agents that run on production Cisco and Genesys infrastructure. Also owns the knowledge layer: RAG pipelines that stay current with live SharePoint and ServiceNow data so agents always have the right context.
Graymatics
Apr 2023 – Nov 2024AI Engineer
Took video intelligence from prototype to production - GPU pipelines processing 40+ concurrent streams on NVIDIA Deepstream, TensorRT-optimized to hit 117 QPS on YOLOv7. Also brought the same models to edge hardware via FP16 quantization on Jetson Nano.
Ensuredit
Apr 2022 – Aug 2022Data Science Intern
Fine-tuned document understanding models on insurance policy PDFs and built an rPPG system that reads vitals - heart rate and blood pressure - from nothing but a webcam feed.
NIT Trichy
Oct 2021 – Feb 2022Student Research Intern
Researched travel-time prediction using GNNs on road network graphs. First real exposure to applying ML on structured graph data - also built autoencoder pipelines for feature extraction on traffic datasets.
Projects
OpenClaw SaaS
2026Managed SaaS that lets users run OpenClaw on cloud instead of their own machines. Container-per-tenant isolation on AWS ECS Fargate, wildcard ALB routing for personal subdomains, and EventBridge + Lambda for real-time container state sync.
LLM Gateway
2025Enterprise AI chat platform with document-grounded RAG. All LLM requests route through LiteLLM Proxy for provider abstraction and rate limiting. Celery workers handle async document processing, Qdrant for vector search, and Langfuse + Phoenix for observability.
Virtual Try-On
2024Implemented Virtual Try-On using diffusion model checkpoints and CLIP prompt tuning; matched competitive baseline results.
About
I build AI systems that ship - agentic workflows, RAG pipelines, GPU inference infrastructure, and the glue that holds them together in production. My work centers on closing the gap between what AI can do in a notebook and what it takes to run reliably at enterprise scale.
Currently at FuturePath AI, building autonomous agents and real-time AI products for Fortune 500 clients. Previously at Graymatics, building video intelligence infrastructure running 40+ concurrent streams on NVIDIA Deepstream.
Capabilities
- •Voice AI & Real-Time Systems
- •RAG & Knowledge Pipelines
- •GPU Inference at Scale
- •Agentic Systems & Orchestration
- •LLM Infrastructure & Multi-Provider Routing
- •Edge Deployment (Jetson, TensorRT)
Tools
- •Python, TypeScript, C++, CUDA
- •LlamaIndex, LangChain, LangGraph, DSPy
- •LiveKit, Cisco/Genesys Telephony
- •NVIDIA Deepstream, Triton, TensorRT, ONNX
- •FastAPI, Celery, Prisma
- •Docker, Kubernetes (EKS/GKE/AKS), GitHub Actions
- •OpenAI, Azure OpenAI, AWS Bedrock, Vertex AI
- •PostgreSQL, pgvector, Qdrant, Redis
Education
2023
IIIT Naya Raipur
B.Tech, Electronics & Communication Engineering
CGPA 8.25
Contact
Let's build something.
Currently available for full-time roles and consulting projects.
© 2026 Pradyumna Yadav