Devansh Bhagania

Python, FastAPI, PostgreSQL, Redis, pgvector, Hugging Face, Playwright · Aug 2026 – Present

• Built the platform largely solo; it serves 3,000+ users and 30+ employers who moved their recruiting fully onto its ATS.
• Built AI sourcing agents that read a role, find matching candidates, run outreach, screen replies and book interviews, achieving a 28% candidate reply rate.
• Built RAG pipelines on pgvector over domain-specific knowledge bases for context-aware, grounded AI responses, and fine-tuned Hugging Face conversational models per client.
• Built an open-source job scraper (Flask, Playwright) that pulls recent listings from 8 portals, including LinkedIn, Indeed, Naukri and Glassdoor, handling bot-protected sites with persistent browser sessions.
• Designed the PostgreSQL schema, indexing and query tuning behind search, matching and the ATS pipeline.

Python, FastAPI, Flask, Pandas, NumPy, BSE StAR MF API, WhatsApp Business API · Jul 2026 – Sep 2026

• Integrated the BSE StAR MF API end to end for lumpsum and SIP order placement, with idempotent APIs that prevent duplicate orders; the platform manages ₹4 Cr+ in AUM.
• Built AI portfolio analysis with scheduled cron jobs that score and track each portfolio on risk, consistency and performance.
• Built high-throughput Python services (FastAPI, Flask) with Pandas and NumPy data pipelines, and WhatsApp Business API notifications for orders and portfolio alerts.
• Owned high- and low-level system design for modularity, scalability and security.

MedusaJS, PostgreSQL, Redis, ComfyUI, Docker, AWS ECS Fargate, GitLab CI · Oct 2026 – Present

• Built an AI image generation pipeline in ComfyUI for product imagery, contributing to a 60%+ increase in customer conversion.
• Built AI agents for vendor operations that automate catalogue management and SEO across 5,000+ SKUs.
• Integrated AI-driven workflows into the marketplace to streamline vendor operations and improve product discoverability.
• Built a bulk upload system (validation, batching, S3 ingestion) and GitLab CI zero-downtime deploys on AWS ECS Fargate with Dockerised services.

• Workspace that connects docs, API references, architecture decisions, code context and tasks, so every task arrives with what to build, why and where.
• Docs in Markdown with Mermaid diagrams and code highlighting, plus Postman collection import for API reference with environments and tokens.
• Tasks carry their reasoning, acceptance criteria, dependencies and two-way links to related docs, files and endpoints.
• MCP server lets coding agents such as Claude Code, Cursor and VS Code publish and update docs with the permissions of the user who connects them.
• Server-side tenant isolation, six built-in roles plus custom roles with 20+ permissions, project-level access, expiring or revocable share links and activity history.