PxA Labs. A research collective building open, measurable intelligent systems, from safety monitors for fine-tuned LLMs and accent-robust speech recognition to interpretable genomics and real-time adaptive optics.
PxA Labs is an independent, multi-disciplinary research collective based in India. It was founded by two engineers who design, build and maintain almost every project in the organisation themselves. We take a research question from idea to working system: a reproducible notebook, an open model, a tool people can install. Every result is reported against a fair baseline, including what did not work.
| Area | Focus | Start with | |
|---|---|---|---|
| 01 | AI Safety & Alignment | Monitoring and enforcing safe behaviour in models and agents, from refusal circuits to deterministic policy engines. | safetypulse · Mandate |
| 02 | Language & Speech | ASR and auditory interfaces that close regional accent gaps for Indian-accented speech. | IndiVoice-DeepASR |
| 03 | Bioinformatics | Interpretable machine learning that decodes regulatory signals in non-coding DNA on modest compute. | interpretable-regulatory-genomics |
| 04 | Multi-Agent Systems | Agent teams that act inside real tooling, with observability and runtime validation built in. | AgentsScope · AutoMaintainer |
| 05 | Signals, Optics & Systems | Real-time adaptive optics, acoustic physical layers and fault-tolerant data infrastructure. | Project-RIPRA · ultrasonic-phy-layer · ReliaDL |
- Honest baselines. Every claim is measured against a fair baseline and reported as-is.
- Real compute budgets. We design for the hardware people actually have: Kaggle T4s, Colab, consumer GPUs.
- Open by default. Code, notebooks and models are published so others can reproduce, audit and extend them.
- Systems, not demos. Checkpointing, observability, audit trails and fault tolerance from day one.
We welcome research collaborations, contributions and conversations. Open an issue on any repository, join the Discord, or write to pxalabs@gmail.com.
PxA Labs. Research and engineering for intelligent systems.
