R&D Intern @ Alippo Tweets about his journey ,Peace lies in solidity Github:https://xitter.cf/t.co/oOdSJdK2Al
Joined December 2019
- Tweets437
- Following78
- Followers70
- Likes510
Pinned Tweet
Won Central's India largest hackathon, @HackbyteTPC
Last time, I wasn't even selected for final round. Well this time, I came prepared 😎
Built an AI mock interviewer platform which provides guidance, DSA and system design interviews
#Hackathon #hackbyte #GenAI
Sending emails from Google Sheets has always been a challenging task for any college club. I built this in the last two days to automate the process. Plus, plain emails look dull and unattractive. Manually adding each person? Ehh, no thanks!
After hours of procrastinating, the LeetBot v2 is almost complete. I finally switched to Phidata from Lang-Graph due to prior experience.
Phidata is pretty good for these tasks, but it lacked Supabase support for both agent storage and memory. I finally added this myself.
Integrated LangGraph with the current implementation
Three things more:
1) A database to store and record user queries and chat history- thinking of implementation like openai threads
2) Scraper cron job initialization
3) initialise remaining hints and similar question agent
LeetBot v2 updates:
The General_Query agent is done. It was the tough to implement because it had to use two-tools
I explored NeMo Guardrails and Guardrails AI for guardrailing the response, but they didn't support GROQ. Proper prompting was necessary to keep it on topic
The two tools which I have implemented are:
1) DSA classifier to classify topics
2) DuckDuckGo search for proper and detailed answering
For prompt-guarding, I explored nemo-guardrails and guardrails ai but they didnt support groq properly and also it added latency .
LeetBot-V2 updates:
Implement the query router agent: Now the agent can successfully differentiate between queries
Implemented the Github agent: Now agent is capable of uploading your code directly to Github itself!
2) Now comes the fun part, implementing github api agent. Well to be honest, it wasnt easy .I tried to implement this from scratch but deliberately failed to do so, had to take help from langchain's agent and tool system.
A high level overview of LeetBot V2. Plan is pretty simple and clear .
I'll be using LangGraph to execute this agentic workflow.
Never used LangGraph though, firstly would be making a Corrective Rag agent to get familiar with it.
Vikram retweeted
A simple fact, if you think some section of society is special, then you'll never reach their level.
" 3 digit IIT rankers are special"
"Some engineers are built different, they are 10x, I can't compete with them"
You'll never match them not because of the skill gap, but because you already accepted you can't in your mind.
If you see them for what they are "people just like you who put in the hours, taken 1000s of shots & iterated" you can see a path, by calling them special or lucky you are just disrespecting all the hours & efforts them have put in.
Mindset -> choices -> actions.
Vikram retweeted
💛🖤👋🏻 Marco Reus’ last game as BVB player will be the Champions League final at Wembley.
Day 1 of breaking out of TUTORIAL HELL
Building a better version of ${originalNameToBeRevealedlater} , Planning done . Now its time for execution
Let's see how far I can go to complete this admist my current end-sems and internship
