The Future of AI Agents - Aditya Gautam

9 snips

Mar 6, 2026

Aditya Gautam, AI researcher and engineer with experience at Google and Meta, specializing in LLMs, recommender systems and agent architectures. He discusses enterprise adoption hurdles, the economics of fine-tuning versus APIs, agent MLOps needs like guardrails and data lineage, reliability in regulated sectors, designing multi-tenant evals, human-in-the-loop workflows, and the future of multimodal autonomous agents.

Ask episode

AI Snips

Chapters

Transcript

Episode notes

INSIGHT

Enterprise Adoption Follows Three Clear Steps

Enterprises face three adoption hurdles: understanding AI, integrating it into legacy systems, and monitoring/improving deployed agents.
Aditya learned this by interviewing dozens of small businesses and investors who struggle with legacy tools and third‑party integrations.

ANECDOTE

Legal Bots Lost Some Users To General Chatbots

Harvey built legal vertical tooling and initially gained lawyer adoption, but general chatbots improved and some users migrated away.
Aditya heard VCs say lawyers started preferring general chatbots like Gemini for many queries over niche legal bots.

INSIGHT

Agents Need Governance To Prevent Hidden Collusion

Multi‑agent reliability needs governance: audit trails, policy enforcement, and detection of collusion between agents.
Aditya warns agents can leak combined signals to a third agent, requiring lineage and governance infrastructure.

Get the Snipd Podcast app to discover more snips from this episode

Get the app

In this talk, Aditya, an experienced AI Researcher and Engineer, shares his technical evolution—from his roots in embedded systems to building complex, large-scale AI agent architectures. We explore the practical challenges of enterprise AI adoption, the shifting economics of LLMs, and the infrastructure required to deploy reliable multi-agent systems.You’ll learn about:- The ROI of Fine-Tuning: How to decide between specialized small models and general-purpose APIs based on cost and latency.- Agent MLOps Stack: The essential roles of guardrails, data lineage, and auditability in AI workflows.- Reliability in High-Stakes Verticals: Navigating the unique AI deployment challenges in the legal and healthcare sectors.- Evaluation Frameworks: How to design robust evals for multi-tenancy systems at scale.- Human-in-the-Loop: Strategies for aligning "LLM as a judge" with human-labeled ground truth to eliminate bias.- The Future of AGI: What to expect from the next wave of multimodal agents and autonomous systems.TIMECODES: 00:00 Aditya’s from embedded systems to AI08:52 Enterprise AI research and adoption gaps 13:13 AI reliability in legal and healthcare 19:16 Specialized models and agent governance 24:58 LLM economics: Fine-tuning vs. API ROI 30:26 Agent MLOps: Guardrails and data lineage 36:55 Iterating on agents with user feedback 43:30 AI evals for multi-tenancy and scale 50:18 Aligning LLM judges with human labels 56:40 Agent infrastructure and deployment risks 1:02:35 Future of AGI and multimodal agentsThis talk is designed for Machine Learning Engineers, Data Scientists, and Technical Product Managers who are moving beyond AI prototypes and into production-grade agentic workflows. It is especially relevant for those working in regulated industries or managing high-volume API budgets.Connect with Aditya:- Linkedin - https://www.linkedin.com/in/aditya-gautam-68233a30/Connect with DataTalks.Club:- Join the community - https://datatalks.club/slack.html- Subscribe to our Google calendar to have all our events in your calendar - https://calendar.google.com/calendar/r?cid=ZjhxaWRqbnEwamhzY3A4ODA5azFlZ2hzNjBAZ3JvdXAuY2FsZW5kYXIuZ29vZ2xlLmNvbQ- Check other upcoming events - https://lu.ma/dtc-events- GitHub: https://github.com/DataTalksClub- LinkedIn - https://www.linkedin.com/company/datatalks-club/ - Twitter - https://twitter.com/DataTalksClub - Website - https://datatalks.club/