Karthikeya Kollu
AI Engineer
Overview
Social Links
About
My name is Karthikeya. I like working with small-sized teams, preferably less than 30 people. I usually work around 15 hours daily on my side projects as well as my main work. I enjoy helping the community and contributing to open source projects.
I have around 1.5+ years of experience in development and 1+ year of professional experience building products. I like exploring how I can make scalable systems, like RAG pipelines and agentic workflows with task delegation.
I launched my own Chrome extension startup. Within 3 months, it reached around 500 users, and to this date, it has around 1,000 users.
I enjoy solving interesting problems and researching new things.I focus on implementing new features or researching innovative solutions.
GitHub Contributions
Stack
Experience
MyAcolyte EdTech
Current Employer- Redesigned platform for scalable, real-time AI workflows.
- Built a custom agentic AI library for Acolyte AI, inspired by AgnoAGI but with original strategies and multi-agent task delegation.
- Deployed production-ready AI agents with live monitoring, RAG, tool use, and multi-agent delegation.
- Reduced Docker container sizes by 38%, improving deployment speed and resource efficiency.
- Designed efficient fingerprinting algorithm for duplicate detection and linking in a central doc repository to optimize RAG.
- Developed a robust Examination Generation system: reduced question-generation time for 100 NEET-level medical questions from 30 minutes to 500 seconds (without parallelism, via optimized data/model flows).
- Refactored RHE flow: added permission and chunk tables, removing duplication; checked chunk existence to streamline retrieval and access control.
- Implemented dual-study agents: Smart Study Mode (ChatGPT-like) and Intensive Thinking Mode, enabling intelligent role switching for better study assistance.
- Built a PDF viewer with annotation support. leading to recurring API licensing deals (three-figure monthly revenue).
- Maintained project organization on ClickUp, ensuring smooth coordination.
- Managed 3 interns + 1 full-time dev.
- Interviewed over 30 experienced engineers, supporting technical hiring and team scaling.
- TypeScript
- Next.js
- React
- TailwindCSS
- Python
- AWS
- Docker
- RAG Systems
- Vector DB
- Agentic AI
- Team Leadership
- Recruiting
- ClickUp
Orian (Ollama WebUI)
Current EmployerAgnoAGI Framework
Mem0
Open Source <Partial>
Education
Projects(6)
An LLM-based OS that performs tasks according to prompts:
- Full system access, allowing modification of system settings and files.
- Replans and adapts efficiently if initial attempts fail (80%-95% success rate).
- Completely local, ensuring data privacy with no external data exposure.
- Python
- OpenAI
- Ollama
Gumloop is a DOM extension designed to replace Gummy:
- Creates, edits, and modifies flows fully from scratch based on user prompts.
- Automates tasks in the browser efficiently.
- Maximum flexibility for AI-driven web workflows. Additional reference: Day 4 Notes
- JavaScript
- DOM Manipulation
- AI Flow Automation
- Web Extensions
- Prompt Engineering
AI tool integrated with Gemini and Ollama models that enhances:
- Email writing, content summarization, and coding assistance.
- Boosts efficiency by over 30% during device use.
- Supports skill development and coding education for learners.
- Python
- Tkinter
- Ollama
- LangChain
- Gemini API
Workflow Builder application:
- Users can visually construct workflows with drag-and-drop interface.
- Save, load, and execute workflows for data manipulation tasks.
- Backend interprets and executes tasks efficiently in sequence.
- Python
- Flask
- ReactJS
- ShadcnUI
- Firebase
- ReactFlow