Thankyou @kirat_tw for featuring me on the channel. Its a pod where I shared my journey of being selected for GSOC(Google summer of code). Go checkout and don't forget to like!
Thankyou @openWISP for selecting me๐
TIMESTAMP: 56:18
https://t.co/EoRzvlN762
Commiting to learn and build sandboxing platform from scratch like e2b/judge0/daytona.
Will share each day my progress till I build the sandbox and get 100 users on it using actively !!
By the end I be wouls have learned:
Linux namespaces
Cgroups
Containerd
OCI runtime spec
Comtainer runtimes
Docker internals
Sandbox lifecycle
Worker scheduling
And more...
Step-By-Step LLM Engineering Projects Roadmap
- Build a tokenizer
- Learn embeddings
- Implement RoPE / ALiBi
- Hand-wire attention
- Build MHA
- Build a Transformer block
- Train a mini-former
- Compare objectives
- Build sampling
- Speculative decoding
- KV cache
- MQA / GQA / MLA
- Long context
- FlashAttention
- Hardware budgets
- Toy MoE
- Sparse model trade-offs
- State-space / linear attention
- Diffusion language models
- Data pipelines
- Synthetic data
- Scaling laws
- SFT / DPO / RLHF / GRPO
- Quantization
- Serving stacks
- Eval harnesses
- RAG
- Tool use / agents
- Vision-language adapters
- Interpretability
- Red-team suite
- Full capstone model system
One request:
Choose an Opensource AI lab when you make it
Opensource is where humanity gets to keep the tools
DM me when you've made it ;)
Asking my own agent to create a tool to expand itself so it can be more powerful agent.
Aka chitti creating its own replicas... hehe
Now i will use pushpit-agent instead of codex or claude code :)
Am i the only one who feels reviewing ai generated code can sometimes be non enjoyable process?
Then i start thinking, i could have written it myself OR spend moree time in specs๐