LangGraphDeepAgentsFastAPINext.js
Atlas — AI Editorial Agent
Eight-stage LangGraph + DeepAgents pipeline for long-form publishing: research, outline, draft, review, improve, and human approval gates, with persistent skill memory.
Generative AI Engineer — Agents, RAG & ML Systems - Toronto, Canada
Building practical AI systems that move from prototype to product.
I'm Abhay Raj Singh, a Generative AI engineer with 9+ years building production ML and agentic AI systems — from document parsing and intelligence platforms to real-time voice avatars. The projects below are what I build outside the day job: agent workflows, RAG pipelines, and full-stack AI products in Python, TypeScript, LangGraph, FastAPI, Next.js, and Supabase.
01
Agent workflows
Planning, memory, tool use, orchestration, and human approval loops.
02
Applied GenAI
Research assistants, content systems, travel planners, voice agents, and automation tools.
03
ML foundations
Computer vision, deep learning, ML interview preparation, and project-based education.
Selected Work
Recent AI products and developer platforms
Next.jsLangGraphFastAPIWebSocket
AI Agent City Game
Playable city simulation where autonomous citizens have private memory, run LangGraph exchange nodes, and build relationships over time.
Next.jsFastAPILangGraphPostgreSQL
AI Trip Planner Agent
Tool-first travel concierge: intent parsing, provider calls, route optimization, and a repair loop before returning a structured, source-cited itinerary.
Next.jsLangGraphSupabaseClerk
Flowpost
AI social media automation platform: content discovery, image generation, and cross-channel scheduling with a self-improving engagement feedback loop.
Next.jsDeepAgentsTypeScriptZustand
Product Dev Blueprint
Schema-first planning app that pressure-tests an idea and generates a traceable PRD, SOW, system design, and coding-agent prompt bundle.
FlaskTwilioSupabasepgvector
WhatsApp Invoice Assistant
Receipt and invoice workspace over WhatsApp: structured extraction, human-in-the-loop approval, agentic RAG spend Q&A, and outgoing invoice generation.
Stable DiffusionSadTalkerGFPGANFlask
Digital Human RAG Agent
Generates a talking "digital human" video from a character description and script, combining local LLMs, image generation, TTS, and lip-sync video.
LangGraphFAISSOllamafastrtc
Customer Support RAG Voice Agent
Real-time voice support agent running fully local models: speech-to-text, FAISS retrieval over a private knowledge base, and text-to-speech response.
LangGraphFastAPIReactLangFuse
Docugen
Multi-agent SaaS that turns a research query into a professional report, presentation, or manuscript via deep search and conditional LangGraph routing.
Next.jsSupabaseOpenAI RealtimeClerk
ML Interview Roadmap
133-day interview-prep platform spanning ML, GenAI, LLMOps, and system design, with a deepagents-style AI Tutor for chat and realtime voice coaching.
LangGraphLangChainChromaDBLangSmith
OmniAssist
Hierarchical multi-agent personal assistant: a supervisor agent delegates email, calendar, Slack, research, and document RAG tasks to sub-agents over chat.
LangGraphFastAPINext.jsOpenAI
AI Portfolio Agent
The meta project: an agentic command center that builds a developer's portfolio from GitHub and a resume, then drafts READMEs and opens real pull requests.
Toolbox
Stack shaped around AI delivery
AI and ML
Deep learning, machine learning, computer vision, RAG, vector search, agent workflows, prompt systems.
Backend
Python, FastAPI, LangGraph, CrewAI, Supabase, pgvector, API design, workflow orchestration.
Frontend
TypeScript, JavaScript, React, Next.js, product dashboards, developer-facing tools, responsive interfaces.
Delivery
Vercel, GitHub Actions, production scaffolds, documentation, automation, and practical product iteration.
Contact