agent/data-collectorv2.1
Multi-agent ingestion for LLM training — search, extract, transform, verify agents with retry logic and error isolation. Milvus + Neon.
pipeline: scrape → chunk → embedsource ↗
traces / ashwin.system / root span
Ashwin H
AI engineer building RAG systems, multi-agent architectures and LLM infrastructure in production. This site is the system itself — every claim resolves to a source, and the lab runs the real mechanics in your browser.
span 01
experience.waterfall
Jan 2022 → now · durations to scale2022
2023
2024
2025
2026
jagriq· AI Engineer
igit.in· Co-Founder & Technical Lead
freelance· Full-Stack & Cloud
7mo
ciragu· Autonomous Systems (Gov.)
1y 3mo
consultadd· Software Engineer
1y 1mo
ford· DBA Intern
6mo
span 02
agents.registry
personal projects, running as servicesagent/neurovaultv1.4
RAG-based memory engine — multi-document ingestion, semantic chunking, Milvus + Chroma indexes, sub-second contextual recall.
p50 recall < 1ssource ↗
agent/learnxv1.0
Multi-agent teaching platform — Planner and Mastery agents on LangGraph, persistent memory on Neon, difficulty tuned by live signals.
agents: planner · masterysource ↗
span 03
system.metrics
numbers from production, not adjectives0K+
documents in RAG corpus
0K
structured units indexed
0K+
entity relationships
0
languages, OCR ingestion
0+ yrs
shipping in production
0.0%
uptime, client platforms
span 04
stack.manifest
B.Tech CSE, SRM Chennai (85%) · 4+ yrs in productionai / ml
cloud / infra
full-stack
vectors / db
span 05
contact.endpoint
open channel · low latency repliesPOST /contact HTTP/1.1
host: ashwin.system
content-type: application/json
{
"email": "ashwin21rr@gmail.com",
"github": "github.com/ashwinaadithan",
"linkedin": "linkedin.com/in/ashwinaadithan",
"phone": "+91 93600 84775"
}
HTTP/1.1 202 Accepted // I read everything. Ship me something hard.