Conf42 Prompt Engineering 2024
2024-11-14
-- online --
who it's for
prompt engineers, ai engineers, ml engineers, mlops engineers, software engineers, site reliability engineers, security engineers
topics
prompt injection defense; agent orchestration for super agents; real-time network failure forecasting; GitHub Copilot and code generation; retail demand forecasting with pandas; vector search and embeddings; cloud cost optimization and FinOps; on-device large language models; OCR with LLMs for structured data; reducing hallucinations in conversational systems; testing LLM-powered applications; GPTScript and prompt-driven automation
| title | author | length |
|---|---|---|
| Model-Based Input Validation: Preventing Prompt Injection Attacks | Hilik Paz | 17m |
Claims a practical defense: a meta prompt runs before the LLM to classify inputs, returning a pass/fail JSON plus a secret token that detects tampering… | ||
| AI Orchestration: Getting Super Agents to Complete a Mission | Batandwa Mgutsi | 25m |
Claim: orchestration can scale to 100 people with one human by composing agents. Demo shows a single‑layer agent (ChatGPT 3.5 + voice service) and a… | ||
| Predictive Network: How AI Forecasts System Failures | Akshat Kapoor | 18m |
Claims that AI models using LSTM and random forests plus real-time packet logs flow telemetry and device configs predict link OS and human-error faults with up… | ||
| GepAI in Development Processes and Dev Tools | Bohdan Snisar | 22m |
Claims that large language models identify gaps in requirements and turn user feedback into ready tickets, demonstrate concrete use of GitHub copilot and… | ||
| Optimizing Retail Forecasting Through AI Models | Sijo Valayakkad Manikandan | 38m |
Claims that generative AI speeds proof‑of‑concepts and code generation for retail forecasting, producing Python scripts (pandas, statsmodels, matplotlib), EDA… | ||
| AI-Driven Assessments for Continuous Learning | Vijay Kumar Valaboju | 12m |
Claims a working reinforcement-learning pipeline that drives real-time adaptive assessments and dynamic curriculum adjustment, with algorithms that change… | ||
| Unlocking the Future of HR Tech | Marina Goplacheva | 17m |
Argues that conversational AI reduces change management by letting employees use their own words and by integrating into messaging apps and existing enterprise… | ||
| Mastering the Art and Science of AI Communication | Charlie Reynolds | 10m |
Argues that concrete pipelines matter: stringing together chat BT, a sentiment analyzer and a summarizer to process inputs yields different outcomes than a… | ||
| How you can trick your company into hiring ChatGPT | Daniel Pries | 41m |
Claims that many automated coding screens are solvable by ChatGPT 3.5 and Co‑pilot, uses Chet Gupta — an AI persona with Dolly images and synthesized… | ||
| End-to-End Chatbot Web App with Your Own Data | Konstantinos Passadis | 47m |
A step‑by‑step build of an end‑to‑end chatbot that actually ingests PDFs, creates embeddings with an OpenAI deployment, registers a vector index in Azure AI… | ||
| Optimizing Cloud Efficiency with Prompt Engineering | Hari Yerramsetty | 34m |
Claims 30% of cloud spend is wasted, demonstrates use of AWS Trusted Advisor and Cost Explorer for visibility, and shows concrete levers: right‑sizing… | ||
| Cloud Technologies for Intelligent Healthcare Solution | Praveen Valaboju | 08m |
CI/CD pipelines and containers cut deployment faults by about 50% fewer releases with errors, while Kubernetes autoscaling keeps services up during spikes and… | ||
| Optimizing IT Performance in Complex Systems | Ravitej Veeramachaneni | 21m |
Performance remains a moving target: complexity of microservices, cloud and IoT causes frequent root‑cause ambiguity (a restarted database sometimes “fixes” a… | ||
| Challenges, Considerations, and Best Practices | Aditi Godbole | 13m |
Claims that clear prompts cut cost and hallucinations and that iterative refinement yields reliably usable outputs For example a redesigned brief asks for five… | ||
| How I Prompted AI to Build a Pregnancy App I Could Trust | Rizel Scarlett | 18m |
Claims a working pregnancy web app that keeps user-owned data in Decentralized Web Nodes (DWN) and shows the actual DWN record printed to prove it. Builds a… | ||
| Exploring Copilot Capabilities | Peter De Tender | 45m |
Live demos show copilot.microsoft.com switching between web and work contexts, M365 copilot pulling SharePoint/OneDrive docs, and PowerPoint auto‑building… | ||
| Reducing Hallucinations in Conversational Systems | Parvin Gasimzade | 15m |
Claims that prompt techniques cut hallucinations without retraining sit alongside concrete numbers: GPT-3 has 175 billion parameters. Gives specific tools and… | ||
| Techniques for Enhanced Model Performance | Reetha Vadakkekkara | 18m |
Claims that prompt engineering measurably raises model usefulness and gives concrete techniques: use GPT-4 and image model DALL·E alongside ChatGPT and… | ||
| Real-Time Data for Generative Feedback Loop (GFL) Automation | Bobur Umurzokov | 28m |
Concrete pipeline claim: a real‑time generative feedback loop that detects Supabase row changes, calls OpenAI to generate or enrich listing descriptions… | ||
| Codeless Generative AI Pipelines: Prompt Automation | Timothy J Spann | 12m |
Shows a concrete pipeline: ingest and chunk real time data, produce vector embeddings, store them in Milvus vector DB, run approximate nearest neighbor… | ||
| Inside Multimodal Models | Bethany Jepchumba | 13m |
Explains the actual architecture and hands-on results: attention + shared Transformer layers, encoder fusion and contrastive training that CLIP contrastive… | ||
| The Disruptive Potential of On-Device LLMs | Rishab Mehra | 22m |
Claims that on-device models cut privacy risk and carbon cost and that Llama 3.1 8B and a 70B Llama behave very differently in practice. Demos run a one-line… | ||
| Practical Strategies for Implementing Domain-Specific LLMs | Chaitanya Pathak | 22m |
Claims concrete wins: a deployed RAG based inference engine on AWS that scales behind firewalls, semantic chunking outperforms fixed chunking, and BM25/DFIDF… | ||
| Can LLMs Understand the Linux Kernel? | Yusheng Zheng | 11m |
A survey-driven pipeline uses language models plus agent workflows to label and quantify development activity across the Linux tree, producing a structured… | ||
| Extracting Structured Data from Images with OCR and LLM | Vladimir Pesterev | 13m |
Builds a Deno TypeScript receipt-scanner API that uses Tesseract.js to OCR receipt images, passes the raw text to OpenAI function calling with JSON mode and a… | ||
| Streamlining Property Search Across Multiple Platforms | Kirill Mozheykin | 26m |
Claims that a meta search engine aggregates listings from over 10,000 websites in 30 countries and actually reduces duplicates and fills missing fields using… | ||
| Leveraging ML to Create a Reliable Anti-Fraud System | Pavel Zapolskii | 16m |
Claims a working, production anti-fraud pipeline built over four years and deployed in Yandex advertisement that detects payment and gaming cheats by modeling… | ||
| The Smart Shortcut from Idea to Market-Ready Solution | Sabarinath Viji Selvam | 22m |
Claims that AI can handle roughly 30% of specification drafting and that product roles evolve into a hands‑on product Alchemist within 3–5 years lead the talk… | ||
| Crafting Hyper-Relevant User Experiences | Olanrewaju Oyedele | 28m |
Claims specific ROI numbers (50% increase in user interaction, 200% boost in conversion rates), concrete tools (Adobe Experience Cloud, Google AI, Salesforce… | ||
| A Large Language Model Took Me on Vacation | Jorrik Klijnsma | 12m |
A large language model actually plans a Porto holiday end to end — itinerary, directions to a bakery, and a packing list with formatting that improves… | ||
| Prompt Engineering Simplified | Dan Cleary | 13m |
Claims that small prompt tweaks change outputs drastically and that prompt engineering remains one of three main levers for better LLM results. Empirical… | ||
| Drug Launch Success Through Prompt-Driven Sales Data Analytics | Harpreet Singh | 11m |
Claims concrete outcomes and tools: predictive models deliver 80–90% model accuracy, territory optimization yields ~30% efficiency gain, early conversion of… | ||
| An Art, a Science, or Your Next Job Title? | Maxim Salnikov | 41m |
Claims that generative AI already has over 1 billion users, shows a concrete marketing automation prompt that returns a JSON object for downstream processing… | ||
| AI-Driven Prompts for Enhanced System Reliability | Nagarjuna Malladi | 17m |
Practical SRE takeaways with concrete numbers and tools not fluffy buzzwords The talk claims e-commerce must hit page loads under 3s because a 1 second = 7%… | ||
| From Prompts to Automation: GPTScript and Fabric | Piotr Zaniewski | 29m |
Claims that command-line prompt automation speeds tasks by combining Fabric patterns and GPT Script with small Unix tools. Demonstrates a working orchestrator… | ||
| AI Workflows with Kafka and Modern Architectures | SreenivasaRao Jagarlamudi | 19m |
Shows a working stack: Apache Kafka with Streams API and ksql handling millions of messages with end to end latencies claimed under 10 ms, Kafka Connect moving… | ||
| Securing IoT Integrations with Third-Party Services | Pavan Vovveti | 11m |
Claims concrete results from a smart city deployment with 20,000 devices and measured outcomes — 20% reduction in pollution and 25% fewer parking incidents —… | ||
| Testing LLM-Powered Applications | Tomas Fernandez | 19m |
Case study of Microsoft's Tay shows a chatbot goes off the rails in 42 hours, and a resume-screener demo reveals prompt injection via white text hidden in PDFs… | ||
| Enhancing AI-Driven Quality Assurance | Ludovico Besana | 27m |
Claims that prompt engineering measurably increases accuracy, speed, and reliability of AI-driven QA by turning test-case specs into repeatable prompts… | ||
| Ethical Dimensions of AI in Educational Technology | SantoshKumar Pulijala | 15m |
Claims and concrete numbers drive the decision: the sector is valued at nearly $90 billion (2020) and projects ~$285 billion by 2027, remote learning reaches… | ||
| nothing matches | ||