More Software Engineering Radio episodes
Published 8 Apr 2026
Duration: 48:02
Ambiguity in language and speech, arising from context, phrasing, and incomplete information, poses challenges for AI systems due to their limited context processing, while humans resolve it through contextual cues, tone, and prior knowledge, with strategies focusing on contextual prompts, audio training, data augmentation, and balancing AI efficiency with human-like adaptability in multilingual and ethical contexts.
Sahaj Garg, co-founder and CTO of Wispr, a voice-to-text AI that turns speech into polished writing, talks with host Amey Ambade about designing syste...
The podcast episode examines the concept of ambiguity in human input, distinguishing it from noise and errors as an inherent property of unclear or multifaceted information. It highlights how humans resolve ambiguity using context, tonal cues, and prior knowledge, while machine learning models face challenges due to limited context windows, which hinder their ability to process and interpret ambiguous inputs effectively. The discussion extends to speech-to-text conversion, where unstructured spoken languagemarked by slang, filler words, and varying formalitiesrequires context-aware processing to adapt to different communication styles and user intent. Key challenges include handling background noise, accents, jargon, and the need for models to leverage contextual information to improve accuracy, especially in voice-first systems like Whisper.
The episode further explores types of ambiguity, such as polysemous words, sentence structure confusion, and stylistic variations in language depending on the audience (e.g., texting vs. professional communication). It addresses the limitations of traditional audio models and the potential of large language models (LLMs) in integrating context, history, and external prompts to enhance speech recognition. Strategies for improving model performance include contextual training with vocal metadata, data augmentation, and refining outputs through instruction tuning aligned with user preferences. The discussion also touches on balancing personalization with consistency, the role of user feedback in refining AI systems, and the importance of context compression and inference optimization in managing ambiguity and ensuring efficient, accurate AI interactions.
2 Jul 2026 Jeroen Janssens and Thijs Nieuwdorp on Using Polars
The *polars* library offers a high-performance, multi-language (Python, R, Node.js) data frame tool leveraging Rust, columnar storage, out-of-core processing, and GPU acceleration for fast data manipulation, optimized for large datasets and scalable data engineering workflows.
24 Jun 2026 Scott Kingsley on the Swagger Ecosystem
The Swagger/OpenAPI ecosystem, including tools like Swagger Editor, UI, and Codegen, along with commercial offerings and contract testing, supports cross-functional API development, evolves from Swagger 1.0/2.0 to OpenAPI under the Linux Foundation, emphasizes language-agnostic JSON/YAML specs, HTTP-centric design, modernization of legacy APIs, balances open-source and commercial tooling, explores JSON/YAML trade-offs, FastAPI integration, contract-first/code-first approaches, AI-driven spec generation, security practices, tooling challenges, governance, linting, mocking, CI/CD validation, and emerging AI-enhanced API trends.
10 Jun 2026 Jure Leskovec on Relational Graph and Foundational Models
Predictive modeling faces challenges with AI's limitations in structured data, prompting solutions like graph databases and relational deep learning with attention mechanisms to enhance accuracy, scalability, and real-time updates for enterprise applications.
3 Jun 2026 Dave Airlie on Linux Kernel Maintenance
The Linux kernel, the largest global software project, uses a hierarchical maintainer system with 80,150 contributors managing subsystems like DRM through public review, structured development cycles, and evolving practices to address scalability, quality, and integration challenges.
27 May 2026 Dwayne McDaniel on the Engineering Challenges of Secrets Management
Managing secrets like credentials and API keys in software development risks leaks causing supply chain attacks (e.g., PyPy, Clot, Cisco) due to secrets sprawl, plaintext storage, and misuse, prompting solutions like time-bound credentials, decentralized systems, vault tools (e.g., HashiCorp Vault), and strategies such as credential rotation and encrypted storage amid over 28.65 million hard-coded secrets in GitHub in 2025.