8000 ayusingh-54 (Ayush Singh) · GitHub
[go: up one dir, main page]

Skip to content
View ayusingh-54's full-sized avatar
🎯
Focusing
🎯
Focusing

Highlights

  • Pro

Block or report ayusingh-54

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. 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
ayusingh-54/README.md

Ayush Singh

AI Engineer at HiLo MRM. Freelance full-stack AI engineer and tech writer. Previously AI Engineer at U2xAI.

I turn language models into working software: agents that call tools, retrieval grounded in real documents,
and pipelines that run on their own. I also write about how they work.

Portfolio LinkedIn Email


Now

  • AI Engineer at HiLo Maritime Risk Management (London, UK · remote). I build ML models for predictive risk assessment and anomaly detection on marine data, LLM-driven decision support, and shipped a hybrid classification system to production on Azure.
  • Freelance Full-Stack AI Engineer & Tech Writer for US startups and tech companies, including Crework and 9otify Technologies.
  • Previously AI Engineer at U2xAI (New Jersey, US · remote): agentic AI for supply chains, including an Oracle MCP server for Claude Code.
  • Open to AI engineering roles and project work. Email is the fastest way to reach me.

What I build

What it means Tools I reach for
agents Multi-step agents that call tools, keep memory and stay inside guardrails LangGraph · CrewAI · AutoGen · MCP
rag Document pipelines, vector search and answers grounded in your own data Pinecone · FAISS · ChromaDB · pgvector
automation Workflows that move data and content between tools without anyone pushing buttons n8n · Zapier · Python · REST APIs
web The interface, API and database around the model, shipped to production Next.js · FastAPI · PostgreSQL
ml Predictive models, anomaly detection and feature pipelines, from notebook to production PyTorch · Scikit-learn · BigQuery · Azure
writing Articles and platform content explaining how AI systems work Substack · long-form · newsletters

Experience

HiLo Maritime Risk Management  London, United Kingdom · Remote

AI Engineer · Feb 2026 – Present

ML and LLM systems for maritime risk, built on vessel, incident and operational data.

  • Built a complete hybrid classification system end to end and deployed it to production on Azure
  • Building machine learning models on marine-industry datasets for predictive risk assessment and anomaly detection across maritime operations
  • Engineering data-processing and feature-engineering pipelines that clean, transform and unify heterogeneous marine data sources for downstream ML and LLM workflows
  • Building a Decision Support System for AI-driven business intelligence and recommendations
  • Improving LLM prompts and agentic workflows for accuracy, consistency and automation across risk-analysis tasks
  • Maintaining requirement traceability across the SDLC, and working with stakeholders to turn maritime-risk business requirements into scalable technical solutions

Machine learning Anomaly detection Feature engineering Hybrid classification Azure LLM-based applications Agentic workflows Prompt engineering LLM evaluation Cost optimisation Decision support Risk modelling

Freelance  United States & worldwide · Remote

Full-Stack AI Engineer & Tech Writer · 2026 – Present

  • Crework (US startup): building automation pipelines and writing articles for the company's Substack
  • 9otify Technologies: built their agentic system and corporate website, then optimised the site so search engines and AI agents can find and use their services
  • Long-form articles and newsletters for other clients' platforms

Automation pipelines Agentic AI SEO / AEO Technical writing

DevsLane (Escecion Technologies)  Noida, India · Remote

Software Engineer Intern · Jun 2026 – Oct 2026
Software Engineer · Apprenticeship · Dec 2025 – May 2026

Full-stack engineer on the in-house ERP that runs HR, finance, projects and hiring, across an AdonisJS backend, a React frontend and a Flutter desktop app. Delivered 20+ features and fixes end to end.

  • WhatsApp leave-management bot for admins on the Meta Cloud API, integrated with the existing ERP
  • Employee ticket desk with per-department keys and least-busy auto-assignment under row locks
  • Encryption at rest with AES-256-GCM, independently rotating key tiers and resumable re-encryption tooling
  • Idempotency-Key support and state guards on 19 API endpoints, plus an approvals Worklist with live socket updates
  • Threaded conversations with @mentions and internal notes on incidents and tasks, shared with the Flutter app

AdonisJS TypeScript React Flutter PostgreSQL Redis WhatsApp Cloud API

U2xAI  New Jersey, United States · Remote

AI Engineer Intern · Dec 2025 – Jun 2026

Agentic AI for consumer-goods supply chains: LangChain and LangGraph agents over retail POS, distributor and inventory data.

  • Engineered an Oracle MCP server for Claude Code so agents can run fault diagnosis and root-cause analysis on Oracle databases, in place of manual DBA inspection
  • Designed BigQuery pipelines for inventory health, demand forecasting and incentive metrics, structured for LLM consumption
  • Built multi-agent LangGraph workflows and custom tool interfaces that let agents reason over supply-chain metrics and recommend actions

LangChain LangGraph MCP Oracle BigQuery Multi-agent systems

Earlier roles
Role Company When Highlights
AI Engineer (intern) Deckoviz Space Labs, UK Mar – Apr 2026 Built Vizzy Chat: GPT-4o and DALL·E 3 orchestrated by a LangGraph state machine behind a Streamlit UI
Full Stack AI Engineer Tycho Technologies, Noida Aug – Dec 2025 End-to-end and agentic RAG pipelines for test monitoring, anomaly detection and QA automation
SDE Intern HapScale, Bangalore Dec 2024 – Jul 2025 Secure API integrations (+25% performance), automated password reset (99.9% fewer manual interventions), +30% landing-page engagement, HapScale Tool tested with 3,000+ people

Selected work

Project What it is Built with
Agent Forge
stars
Open-source collection of 100+ production-ready AI agents for support, research, sales and automation. Started as 100 Days, 100 AI Agents. LangChain · LangGraph · AutoGen · Pinecone · FAISS · Streamlit
OpenClaw A self-hosted Slack AI assistant built from zero: RAG, long-term memory and MCP tool integrations in one hackable codebase TypeScript · Node.js · Claude · MCP · RAG
AI Co-Browser
live ↗
A portfolio assistant that drives the page for you: it navigates, highlights elements and answers questions using OpenAI function calling React · TypeScript · Express · PostgreSQL · Drizzle
Vizzy Chat
Deckoviz Space Labs
One chat interface to create, transform and iterate on visual and narrative content GPT-4o · DALL·E 3 · LangChain · LangGraph · Streamlit
9otify Technologies
live ↗ · client work
Agentic system plus a corporate website, optimised so AI assistants can discover and understand the company's services Next.js · TypeScript · Framer Motion · SEO / AEO
DevsLane ERP
private codebase
Production ERP work: WhatsApp bot, ticket desk, encryption at rest, idempotent APIs, live approvals AdonisJS · React · Flutter · PostgreSQL · Redis
CodeForge-AI A framework-free autonomous coding agent built on ReAct. Cuts routine developer tasks from about 30 minutes to about 30 seconds. Python · GPT-4o · ReAct · MCP · Async IO

Recent builds

A pattern across my 2026 work: agents that show their evidence and stop for a human before anything irreversible happens.

  • Discharge Summary Agent: a planner-driven agent that turns patient source-note PDFs into a discharge-summary draft for clinician review. Every extracted value carries a verbatim quote, and a validator checks that quote against the source.
    LangGraph FastAPI React OpenAI
  • Proposal Copilot: a sales-call transcript goes in and a grounded proposal draft comes out. It retrieves your most relevant past proposals, flags everything a human must confirm and exports PDF or DOCX.
    FastAPI Next.js PostgreSQL + pgvector OpenAI
  • AI Lead Research Agent: researches an inbound lead on the public web, reasons against what it found, and drafts a follow-up with sources attached. There is no send path anywhere in the code, by design.
    Python Pydantic Tavily OpenAI / Anthropic
  • Logos AI · live ↗: a Scripture-grounded assistant on a 9-node LangGraph pipeline with moderation, intent classification, grounding and two-stage output validation.
    Next.js LangGraph OpenAI
  • Meta Ads Audit skill: a Claude Code skill that audits Meta ad accounts with 130+ weighted checks, parallel agents and PDF reports.
    Claude Code Agent Skills
Earlier ML, RAG and data projects

Learning in public

RAG interview guide (105 questions) · BERT fine-tuning, complete guide · MLOps series: CI, DVC, MLflow, Docker, Kubernetes


Stack

Python, TypeScript, PyTorch, TensorFlow, Scikit-learn, FastAPI, React, Next.js, Tailwind, Flutter, Node.js, Express
PostgreSQL, Redis, MongoDB, Docker, Kubernetes, Azure, AWS, Google Cloud, GitHub Actions, Git, Vercel

LLMs & agents LangChain · LangGraph · CrewAI · AutoGen · MCP · ReAct · OpenAI & Anthropic APIs · Claude Code · Prompt engineering
Retrieval & data RAG · Pinecone · FAISS · ChromaDB · pgvector · PostgreSQL · Redis · MongoDB · BigQuery · Oracle
Machine learning Machine learning · Deep learning · NLP · Anomaly detection · Feature engineering · Fine-tuning · PyTorch · TensorFlow · Scikit-learn · Pandas · NumPy
Backend & automation Python · FastAPI · AdonisJS · Express · n8n · Zapier · Socket.IO · Async IO
Frontend React · Next.js · TypeScript · TailwindCSS · Formik · Streamlit · Flutter
Cloud & MLOps Docker · Kubernetes · Azure · AWS · CI/CD · MLflow · DVC · Git

Recognition

  • Techfest, IIT Bombay: hackathon finalist, top 0.5% of 10,000+ applicants in the VISIONX AI competition
  • Smart India Hackathon, internal round: 2nd place among 160+ teams at MMMUT
  • Rise To Top: district-level winner

GitHub activity

Ayush Singh's GitHub stats Most used languages GitHub contribution streak

Tell me what you're building. I'm open to AI engineering roles and project work.
ayusingh693@gmail.com · Portfolio · LinkedIn


Profile views

Pinned Loading

  1. CodeForge-AI CodeForge-AI Public

    CodeForge AI automates all of this using natural language. Describe what you want in plain English, and the AI handles the rest

    Python 3

  2. Agent-Forge Agent-Forge Public

    Welcome to The Ultimate Open-Source AI Agent Collection This is an open-source initiative to create a diverse collection of production-ready AI agents that solve real-world problems

    Python 11 5

  3. Startup-Analysis Startup-Analysis Public

    Indian Startup Ecosystem Analysis & Funding Predictor An interactive dashboard with ML-powered funding prediction for analyzing the Indian startup landscape (2018-2020).

    Jupyter Notebook 4

  4. ironboiiz ironboiiz Public

    Seeking innovative AI models for reflection removal or object detection, targeting real-world applications."[Single Image Reflection Remova

    Python 3

  5. Clinical-Trial-Site-Analysis-using-ClinicalTrials.gov Clinical-Trial-Site-Analysis-using-ClinicalTrials.gov Public

    A comprehensive Python-based system to evaluate clinical trial sites using ClinicalTrials.gov API v2. This system helps identify, analyze, and rank research sites based on multiple performance metr…

    Python 3

  6. AI-Co-Browser AI-Co-Browser Public

    A full-stack AI-powered portfolio website with a co-browsing chat assistant. The chatbot can navigate the page, highlight elements, and answer questions about your projects, skills, and experience …

    TypeScript

0