Skip to content
View AYUSH-148's full-sized avatar
🎯
Focusing
🎯
Focusing

Block or report AYUSH-148

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
AYUSH-148/README.md

Ayush Negi

Backend & AI engineer. Recent IT graduate from ABV-IIIT Gwalior.

Previously interned at CoinSwitch (India's largest crypto platform), building distributed backend services for their F&O trading exchange and security infrastructure. Before that, built real-time conversation intelligence at Profimatix.AI.

What I work on

  • Backend & distributed systems — built 4+ production services handling trade movement, liquidity, and ledger entries at CoinSwitch
  • AI agent infrastructure — contributed to a production MCP gateway giving LLM agents secure access to internal tools (Grafana, OpenSearch, Bitbucket) via OAuth 2.1 & SAML
  • RAG systems & LLM evaluation — built a self-healing RAG agent that reduced hallucination rate from 5.9% → 0.0% with a deterministic eval pipeline and CI gate
  • Voice AI — built with Vapi, ElevenLabs (Flash v2.5, multilingual_v2), and Recall.ai

Featured projects

Project What it does
Self-Healing RAG LangGraph agent that self-critiques and retries its own answers. Hallucination rate 5.9% → 0.0%, with a golden-set eval pipeline in CI. Runs locally on Ollama.
MediFlow Assistant Hybrid GraphRAG medical assistant — vector retrieval + TF-IDF + knowledge graph, merged via Reciprocal Rank Fusion. PII redaction, semantic caching, Langsmith observability.
Threads Full-stack social platform with a write-time notification pipeline: transactional outbox → Redis Streams → SSE. Idempotent delivery, Redis Streams chosen over Kafka based on actual retention needs.

Stack

Languages: Python, Go, TypeScript/JavaScript, C++, SQL

AI/LLM: LangChain, LangGraph, Langsmith, Ollama, Hugging Face, RAG, MCP, Vapi, ElevenLabs, Recall.ai

Backend: Node.js, Express, FastAPI, Flask, gRPC, PostgreSQL, MongoDB, MySQL, Redis, Kafka, RabbitMQ

Frontend: React, Next.js, Tailwind, Redux, Jest/RTL

Cloud & Infra: AWS (EC2, S3, ELB, Lambda), Docker, GitHub Actions

Competitive programming

  • LeetCode Knight — 1900+ rating, 750+ problems solved
  • Global rank 323 — LeetCode Biweekly Contest 135

Links


Building systems that survive production, not just demos.

Pinned Loading

  1. DB-Explorer-MCP DB-Explorer-MCP Public

    MCP server for read-only database intelligence: schema exploration, natural-language queries, EXPLAIN analysis, index suggestions. PostgreSQL / MySQL / SQLite. Zero LLM dependencies.. For remote mc…

    Python 4 1