Alignment is Real // MLOps Podcast #260 with Shiva Bhattacharjee, CTO of TrueLaw Inc.

// Abstract

If the off-the-shelf model can understand and solve a domain-specific task well enough, either your task isn't that nuanced or you have achieved AGI. We discuss when is fine-tuning necessary over prompting and how we have created a loop of sampling - collecting feedback - fine-tuning to create models that seem to perform exceedingly well in domain-specific tasks.

// Bio

20 years of experience in distributed and data-intensive systems spanning work at Apple, Arista Networks, Databricks, and Confluent. Currently CTO at TrueLaw where we provide a framework to fold in user feedback, such as lawyer critiques of a given task, and fold them into proprietary LLM models through fine-tuning mechanics, resulting in 7-10x improvements over the base model.

// MLOps Jobs board

https://mlops.pallet.xyz/jobs

// MLOps Swag/Merch

https://mlops-community.myshopify.com/

// Related Links

Website: www.truelaw.ai

--------------- ✌️Connect With Us ✌️ -------------

Join our slack community: https://go.mlops.community/slack

Follow us on Twitter: @mlopscommunity

Sign up for the next meetup: https://go.mlops.community/register

Catch all episodes, blogs, newsletters, and more: https://mlops.community/

Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/

Connect with Shiva on LinkedIn: https://www.linkedin.com/in/shivabhattacharjee/

Timestamps:

[00:00] Shiva's preferred coffee

[00:58] Takeaways

[01:17] DSPy Implementation

[04:57] Evaluating DSPy risks

[08:13] Community-driven DSPy tool

[12:19] RAG implementation strategies

[17:02] Cost-effective embedding fine-tuning

[18:51] AI infrastructure decision-making

[24:13] Prompt data flow evolution

[26:32] Buy vs build decision

[30:45] Tech stack insights

[38:20] Wrap up