Upgrade to Pro — share decks privately, control downloads, hide ads and more …

LLMOps Operationalizing Large Language Models f...

Avatar for Raj Raj
March 28, 2025
Technology
92
0

LLMOps Operationalizing Large Language Models for the Real World

LLMOps Operationalizing Large Language Models for the Real World

Avatar for Raj

Raj

March 28, 2025

More Decks by Raj

See All by Raj
A Software Engineer’s Guide to Raising Future-Ready Kids
Avatar for Raj rajkumarsakthi
0
170
Power of Minimalism in DevOps
Avatar for Raj rajkumarsakthi
0
150
Minimalism in DevOps 2023
Avatar for Raj rajkumarsakthi
0
180
Are You a Modern Software Engineer?
Avatar for Raj rajkumarsakthi
0
62
Minimalism in DevOps
Avatar for Raj rajkumarsakthi
0
93
Are You are a Modern DevOps Engineer?🤔
Avatar for Raj rajkumarsakthi
0
76
PHP Vegas : Modern Software Developer Best Practices 2020
Avatar for Raj rajkumarsakthi
2
110
PHP & Coding : Modern Software Developer
Avatar for Raj rajkumarsakthi
0
200
Modern Software Developer Best Practices 2020
Avatar for Raj rajkumarsakthi
0
120

Other Decks in Technology

See All in Technology
2026TECHFRESH畢業分享會 - 葬送的通靈師：化系統與用戶雜訊成行動訊號
Avatar for LINE Developers Taiwan line_developers_tw
PRO
0
1.3k
SONiCの統計情報を取得したい
Avatar for SONiC Users Group Japan sonic
0
230
【Cyber-sec+】経営層を"動かす"ための考え方
Avatar for Hisashi Hibino hssh2_bin
0
200
新しいUbuntu/GNOMEが使いたいからXからWaylandへ移行頑張ってるの巻 2026-06-20
Avatar for Nobuto Murata nobutomurata
0
150
IaC コードを資産へ：AWS CDK 社内ライブラリと横断展開 / aws-summit-japan-2026
Avatar for k.goto gotok365
5
940
攻撃者視点で考えるDetection Engineering
Avatar for Ruslan Sayfiev cryptopeg
3
2k
Oracle AI Database@AWS：サービス概要のご紹介
Avatar for oracle4engineer oracle4engineer
PRO
4
3k
FPGAの開発コンペでZephyrを使ってみた
Avatar for misoji engineer iotengineer22
0
130
AWS Security Agent といっしょに脅威モデリングをやってみよう
Avatar for amarelo_n24 amarelo_n24
1
180
スタートアップにAmazon EKSは早すぎる？ マルチプロダクト戦略を加速する Platform Engineeringの実践 / Is Amazon EKS Too Soon for Startups? Practical Platform Engineering to Accelerate a Multi-Product Strategy
Avatar for elmo elmodev09
0
360
Kiroで書いた 設計書 が AI レビューの 採点基準 になる
Avatar for shinji ezaki ezaki
0
130
日本 Fintech 未来予測レポート 2027〜2028年（手動編集版）
Avatar for 8maki 8maki
1
2.5k

Featured

See All Featured
Documentation Writing (for coders)
Avatar for Carmen Chung carmenintech
77
5.4k
Evolving SEO for Evolving Search Engines
Avatar for Ryan Jones ryanjones
0
220
Agile Actions for Facilitating Distributed Teams - ADO2019
Avatar for Mark Kilby mkilby
0
210
Redefining SEO in the New Era of Traffic Generation
Avatar for Szymon Słowik szymonslowik
1
340
From π to Pie charts
Avatar for Rasagy Sharma rasagy
0
210
Self-Hosted WebAssembly Runtime for Runtime-Neutral Checkpoint/Restore in Edge–Cloud Continuum
Avatar for Yuki Nakata chikuwait chikuwait
0
590
The SEO Collaboration Effect
Avatar for Kristina Bergwall kristinabergwall1
1
490
The agentic SEO stack - context over prompts
Avatar for schlessera schlessera
0
820
Making Projects Easy
Avatar for Brett Harned brettharned
120
6.7k
How to Grow Your eCommerce with AI & Automation
Avatar for Katarina Dahlin katarinadahlin
PRO
1
210
Between Models and Reality
Avatar for mayunak mayunak
4
340
30 Presentation Tips
Avatar for Ian Lurie portentint
PRO
1
330

Transcript

  1. LLMOps Operationalizing Large Language Models for the Real World

  2. Hello! I’m Raj 😊 I ❤ to design & develop

    modern software

  3. About me Rajkumar Sakthivel (Raj) Senior Pro Dev @ Tesco

    Technology Founder @ tessellit_uk @rajkumarsakthi

  4. LLMOps Operationalizing Large Language Models for the Real World

  5. AI Disaster

  6. None
  7. None
  8. None
  9. None
  10. None
  11. None

  12. Cost

  13. None

  14. Real-World Challenges

  15. Hallucinations & Safety

  16. Cost Management

  17. Model Drift

  18. Security

  19. Best Practices

  20. RAG Workflow Rule 1: Start with RAG & Fine tune

    later

  21. Prompt Versioning: Rule 2: Treat prompts like code diff tool

    showing how changing ‘Summarize’ to ‘TLDR’ boosted accuracy

  22. GPU Optimization: Rule 3: Right-size GPUs Use A100s for training,

    T4s for inference
  23. None

  24. LLMOps bridges the prototype-to-production gap. Start small, monitor obsessively, and

    never trust an LLM alone with customers.

  25. Any Questions? 😊