Full-stack · AI · India
Devaansh Tyagi
Building AI agents that ship.
Open to freelance,
full-time & collabs.
Work Experience
Aarth.io
Full-Stack & AI Engineering Intern
- Building production full-stack + GenAI features across frontend, backend, and agentic AI workflows
- Architecting LLM features and agent pipelines (retrieval, tool-use, orchestration) in Next.js/Node.js apps with typed APIs
- Implementing enterprise auth: Azure AD SSO (OAuth 2.0 / OIDC), SSL/TLS provisioning, secure session/token management
- Driving DevOps & AgentOps: CI/CD, containerization, prompt versioning, evaluation, agent monitoring, and cross-stack debugging
Project Constellation
Full-Stack Developer · Internal M&A Pipeline Platform
- Architected an internal M&A pipeline platform centralising deal tracking, due-diligence workflows, and stakeholder collaboration
- Built Next.js/React + TypeScript frontend with role-based access, deal-stage dashboards, document management, and workflow views
- Engineered typed backend REST APIs (TypeScript/Node.js), auth, and audit-trail logging with MongoDB across the acquisition lifecycle
- Deployed on a hardened Linux VPS with PM2, Nginx reverse proxy, and Let's Encrypt SSL/TLS for enterprise access
Ernst & Young (EY)
SAP Intern (IT)
- Built a Python analytics solution processing enterprise access-control data across 50+ SAP workflows
- Automated validation, preprocessing, and compliance risk detection with Python/SQL — cut manual audit effort by ~30%
- Delivered dashboards for risk metrics and access anomalies, surfacing violations to compliance teams
Vista Equity Partners · Forage
AI in Action Job Simulation
- Applied prompt engineering and NLP workflows to convert unstructured NPS data into executive-ready insights
- Improved analytical output clarity by 50%+ through structured, AI-driven summarisation of qualitative feedback
ML & Research
Reinforcement learning agent achieving 78% profitable trades. Built with RL algorithms, real-time market data feeds, and portfolio risk management.
Ensemble model (Random Forest + XGBoost) reaching 98.2% accuracy on imbalanced transaction datasets. SMOTE + feature engineering pipeline.
Live Automations
8+ production agents running on n8n — scraping, outreach, market intel, and more.
Web Scraper
Extracts structured data from target sites on schedule
Email Outreach
Personalised cold email sequences with AI-generated copy
Lead Generation
Qualifies and enriches leads from LinkedIn and web sources
Market Intel
Monitors competitor pricing, news, and market signals daily
Content Pipeline
Drafts, formats, and schedules AI-generated content posts
Data Enrichment
Appends company and contact metadata to CRM records
Meeting Scheduler
Parses natural-language requests and books Google Meet slots
Alert Monitor
Watches for business triggers, fires Slack and email alerts
Built with n8n · All agents live in production
Projects
No projects synced yet
Let's build something.
Have a project or role in mind? Reach out.