Podcast
Applied LLM Research & Career Growth: Long-Context Evaluation, Prototyping & Industry Publishing
Open original DataTalks.Club episode
Applied LLM Research & Career Growth: Long-Context Evaluation, Prototyping & Industry Publishing
Original Episode
Use these links for the canonical episode and media sources.
- Open the original DataTalks.Club podcast page
- Watch on YouTube
- Listen on Spotify
- Listen on Apple Podcasts
Episode Overview
How do you evaluate and prototype long-context LLMs in a real-world setting while advancing a career as an applied researcher? In this episode Lavanya Gupta — a Carnegie Mellon Language Technologies Institute alum and Sr. AI/ML Applied Scientist at JPMorgan Chase’s Machine Learning Center of Excellence — walks through practical strategies for applied LLM research and career growth. With 5+ years of industrial research experience, public talks at WiDS, PyData, TensorFlow User Group and reviewer roles for NeurIPS.
People
Use these links to connect the episode to guest notes.
Chapter Summary
Use these checkpoints to decide whether to open the source transcript.
- 0:00 - Episode Introduction & Topic Overview
- 2:02 - Career Overview: From Software Engineering to ML & Master’‘s
- 3:25 - Origin of ML Interest: Hackathons and Computer Vision
- 4:55 - Early Project Case Study: OCR for Organization Charts
- 8:43 - Role Snapshot: LLM Benchmarking at a Financial Institution
- 10:15 - Research Focus: Evaluating Long-Context LLMs
- 12:36 - Empirical Findings: Context Window Performance Droparound 32k–64k
- 14:54 - Practical Approach: Chunking, Retrieval, and Summarization for Large Docs
- 15:28 - Published Work: “Long Context LLMs on Financial Concepts” (EMNLP)
- 17:28 - Industry Research Practices: Publishing from Corporate Teams
- 19:45 - Motivation for Publications: Manager Support and Community Sharing
- 22:10 - Dissemination Paths: arXiv, Endorsement, and Early Publications
- 25:01 - Self-Learning & MLOps: Zoom Camps, Tutorials, and Mentoring
- 30:14 - Rapid Prototyping Tools: Streamlit for Demos and Feedback
- 33:24 - Kaggle Success Story: Building and Licensing a High-Impact Dataset
- 37:32 - Community Contribution: Women in Data Science and Open Mentoring
- 41:13 - Opportunity & Persistence: Timing, Luck, and “Shooting Arrows”
- 45:24 - Career Pivot Guidance: Non-CS Backgrounds Entering Data Roles
- 48:28 - Networking & Mentorship: Cold Outreach and Building Rapport
- 51:28 - Portfolio Strategy: Community Visibility vs. Job-Targeted Projects
- 54:33 - Interview Preparation: LeetCode, Conceptual Mastery, and Mock Interviews
- 56:56 - Project Selection: Industry-Backed Work for Real-World Impact
- 57:46 - Episode Wrap-Up & Final Career Advice