AI Agent Systems
Design and build agentic workflows that connect LLM reasoning with tools, APIs, databases, and human review.
OpenAI Agents SDK, Google ADK, LangGraph, FastAPI, Next.js
AI consulting for founders, product teams, and automation-heavy businesses
Raj Kapadia
AI Agent & Applied LLM Engineer | Founder of TrishiAI | Creator of Calorie Buddy AI
I build production AI agents, conversational systems, and full-stack products—and help teams turn AI prototypes into reliable workflows across WhatsApp, Telegram, web apps, databases, and cloud infrastructure.
7+
years building production software and AI systems
100+
chatbots delivered across Dialogflow ES and CX
5
developers led on AI/ML product delivery
1.2K+
students across applied chatbot courses
Why bring Raj in
Practical AI engineering for teams that need more than a demo.
Raj combines LLM application development, chatbot delivery, full-stack engineering, and applied ML leadership to build systems that can be deployed, maintained, and explained.
Founded TrishiAI for production AI-agent and conversational-product delivery, and created Calorie Buddy AI as a live Telegram-first product.
Led a 5-person AI/ML engineering team building enterprise LLM, text-to-SQL, image search, and deep learning products.
Delivered chatbot and automation projects for clients worldwide through freelance marketplaces and direct consulting.
Brings teaching depth from 6+ years as an Assistant Professor, plus practical production delivery across AI and web stacks.
Team lead delivery
Led AI/ML teams building enterprise LLM, text-to-SQL, image search, and deep learning systems.
Client-ready systems
Delivered chatbots, APIs, automations, and AI products for global freelance and direct clients.
Teaching depth
6+ years as an Assistant Professor plus ongoing YouTube and course content for applied AI builders.
Consulting services
Build AI workflows your team can actually operate.
From prototype cleanup to full product delivery, the focus is reliable agent behavior, practical integrations, and user-facing software that makes AI useful.
LLM Applications
Ship production-ready LLM apps for text-to-SQL, retrieval, internal automation, and decision-support workflows.
RAG, vector databases, prompt systems, evaluations, cloud deployment
Conversational AI
Build reliable web, WhatsApp, and Telegram chatbot experiences with async workers and enterprise webhook patterns.
Dialogflow ES/CX, Meta WhatsApp, Telegram, Redis, ARQ
Full-Stack AI Products
Turn AI prototypes into usable products with durable APIs, dashboards, auth-ready UI, and deployment paths.
Next.js, React, TypeScript, Python, PostgreSQL, GCP, AWS
Ways to work together
Book a fit callPrototype-to-production audit
Review the architecture, failure points, and launch path before you invest in a full rebuild.
End-to-end AI product build
Design and ship the agent logic, APIs, frontend, deployment flow, and handoff documentation.
Webhook and chatbot systems
Build WhatsApp, Telegram, Dialogflow, Gemini, and worker-backed automations that handle real traffic.
LLM workflow rescue
Stabilize unreliable prompts, tool calls, retrieval flows, background jobs, and product UX.
Selected work
Applied AI systems, not slideware.
A sample of agent, chatbot, LLM, NLP, computer vision, and bot products that show how Raj turns model capability into usable software.
AI ProductFeatured
Calorie Buddy AI
Telegram-first nutrition and fitness product for logging meals, photos, voice notes, workouts, and progress without another tracking app.
Problem
Make everyday calorie, macro, and exercise tracking conversational while keeping users in control of what gets saved.
Outcome
A live Telegram product with approval-based logging, personalized goals, daily and long-range summaries, local-time reminders, and an operational admin dashboard.
Text, photo, voice, and workout inputs
Approval before logging
Live product and Telegram bot
TypeScript, OpenAI Agents SDK, Telegram, Next.js, PostgreSQL, Redis, Docker
View buildCreator Tools
YouTube Animation Generator
TypeScript and Remotion CLI for grounded narrated videos and editor-ready animation overlays in landscape and vertical formats.
Problem
Turn source material or subtitles into useful video visuals without fabricating evidence or rebuilding every scene by hand.
Outcome
A deterministic workflow with optional cited research, source-backed charts, local voice synthesis, reusable media, and native 16:9 and 9:16 rendering.
Grounded visual planning
Local narrated-video pipeline
Responsive Remotion rendering
TypeScript, Remotion, OpenAI, Structured Outputs, Local TTS
View buildAgent Workflows
AI Invoice Processing with Human Review
Multimodal invoice extraction backed by deterministic validation, durable LangGraph checkpoints, and reviewer correction.
Problem
Automate invoice processing without allowing model confidence alone to approve financially inconsistent or ambiguous documents.
Outcome
A durable background workflow in which application rules auto-approve clean results and route every warning or mismatch to human review.
Multimodal structured extraction
Durable pause and resume
Deterministic financial checks
Python, FastAPI, LangGraph, Celery, PostgreSQL, MinIO
View buildRAG Systems
YouTube AI Knowledge Base
Dockerized ingestion and RAG backend that turns videos into timestamped, searchable, source-grounded knowledge.
Problem
Make long-form video content searchable and answerable without losing the exact point in the source that supports a result.
Outcome
A background ingestion pipeline with local transcription and embeddings, filtered vector retrieval, and timestamped search and chat sources.
Local Whisper transcription
Timestamped Qdrant retrieval
Grounded answers with source links
Python, FastAPI, Celery, Whisper, Qdrant, PostgreSQL
View buildAI Agents
Same Agent in Three Frameworks
A practical side-by-side reference that makes differences in agent loops, graph routing, and framework discovery concrete.
AI Agents
AI Agents Beyond Jupyter Notebooks
A durable FastAPI and Telegram workflow that processes agent jobs in the background without blocking users.
Conversational AI
WhatsApp + Google Conversational Agents Integration
A production-ready async pattern for Meta webhooks, Dialogflow CX, Gemini, Redis, and background processing.
LLM Applications
LLM-Powered Text-to-SQL WhatsApp Application
A WhatsApp-first text-to-SQL workflow that turns natural language questions into useful database answers.
Applied ML
Toxic Comment Classifier
A practical NLP classifier with a simple interface for testing and demonstrating moderation behavior.
Computer Vision
End-to-End Object Detection Solution
An end-to-end pipeline with training, evaluation, optimization, and API delivery foundations.
Bot Products
Habit Tracker Telegram Bot
A Telegram-native bot with reminders, tracking, and weekly summaries backed by PostgreSQL.
Developer Tools
Dialogflow CX Webhook Template
A reusable FastAPI template that speeds up chatbot implementation and keeps webhook structure consistent.
AI Agents
LangGraph Table Booking AI Agent
A restaurant booking agent concept that demonstrates orchestration across conversation, logic, and UI.
Technical depth
A stack built for AI products that need frontend, backend, and model fluency.
LLM Engineering
Fine-tuning and inferenceAI agents and tool callingPrompt systemsRAG and vector searchOpenAI Agent SDKGoogle ADK
Applied ML
TensorFlowPyTorchComputer visionNLPObject detectionImage similarity search
Product Engineering
Next.js App RouterReact and TypeScriptFastAPINode.js APIsPostgreSQLCloud deployment
Automation Channels
Dialogflow ES/CXWhatsApp integrationsTelegram botsWebhook systemsBackground workersLead capture flows
Teaching and walkthroughs
Recent AI and chatbot tutorials.
Practical implementation videos that show the same applied engineering style used in client work.
Courses
Learn the chatbot foundations behind production consulting work.
Master Google Dialogflow ES: Build Smart Chatbots
Learn to build chatbots using the Dialogflow ES, supported by Python.
70 students
4.6 (16 reviews)
![Master Dialogflow CX - Build Engaging Chatbots [2025]](/_next/image?url=%2F_next%2Fstatic%2Fmedia%2Fdialogflow-cx.0786e5cf.png&w=3840&q=75)
Master Dialogflow CX - Build Engaging Chatbots [2025]
Learn to build chatbots using the Dialogflow CX, supported by Python and FastAPI.
1,206 students
4.2 (33 reviews)
Start a project
Bring a messy AI idea, prototype, or integration problem.
The fastest path is a short call. Use the form if you already have project context, constraints, or links to share.