Naman Jaiswal — AI Engineer
I build production AI systems with strong engineering, clean interfaces, reliable
infrastructure, and real-world deployment discipline. I work across Python, Rust, TypeScript,
cloud platforms, and LLM tooling to ship agentic systems, voice AI, RAG platforms, and AI
observability infrastructure that are fast, practical, and ready for real users. Currently
pursuing B.Tech at NIT Rourkela while building and deploying AI products.
Services & Expertise
- End-to-end AI-native platform development
- Agentic and multi-agent AI systems (LangChain, LangGraph, MCP, A2A)
- Real-time voice AI (Pipecat, Deepgram, ElevenLabs, LiveKit, WebRTC)
- Retrieval-Augmented Generation (RAG) platforms
- LLM tool calling, evaluation, and policy/governance hooks
- AI observability and infrastructure (OpenTelemetry, Prometheus, Grafana, TimescaleDB)
- Workflow automation and production AI delivery
Experience
- AI Engineer Intern — AmericSoft Solutions (Mar 2026 - Present, Remote US):
Building production AI and agentic system components for enterprise clients in regulated
environments. LLM tool-calling, RAG, multi-agent orchestration, observability, and policy hooks.
- AI Developer Intern — ThinkingStone (Jun 2025 - Aug 2025, Remote):
Engineered production voice-agent systems and real-time WebRTC audio infrastructure with
LiveKit, Docker, Fly.io, and Supabase at 99.9% uptime.
Projects
- Agentland — AI agent observability & governance: a black-box flight
recorder for AI agents, one proxy capturing every call, cost, policy event, and audit trace
with sub-5ms p99 overhead. Built in Rust, axum, Tokio, TimescaleDB.
- ShodhAI — Autonomous multi-agent research platform turning a topic into an
expert-grade report via a 7-node LangGraph DAG with parallel analyst agents and live web search.
- RAG Voice AI Agent — Real-time industrial voice assistant with sub-800ms
voice-to-voice latency using Deepgram, Groq, ElevenLabs, Pipecat, and MongoDB Atlas Vector Search.
- VoiceMon — Voice AI observability platform tracking latency, sessions,
errors, and model calls across LiveKit and Pipecat pipelines.
- Ecom Intelligence MCP — Multi-agent e-commerce intelligence assistant for
product research, competitor tracking, and review analysis with RAG and MCP.
Achievements
- Y Combinator Startup School India 2025 — selected cohort, $25,000 AI credits grant.
- NTSE Scholar 2021 — NCERT Top 1,000 out of 1M+ national candidates.
- State Child Scientist 2019 — Dept. of Science & Technology, Govt. of India.
Contact