Metta-AI/metta
Sep 2025 – Oct 2025
2 Months active
Languages Used
C++PythonCUDACudaPyTorchShell
Technical Skills
Backend DevelopmentBug FixingC++Code IntrospectionCommand-line Interface (CLI)Game Development
Subhojeet Pramanik
PROFILE
Subhojeet Pramanik
Subhojeet Pramanik contributed to the Metta-AI/metta and Metta-AI/mettagrid repositories by developing modular memory architectures and enhancing agent training workflows. He built Cortex, a GPU-accelerated library in Python and C++ for recurrent neural networks, integrating AxonCell and AxonLayer to support meta-recurrence and scalable sequence modeling. In MettaGrid, he introduced soul-bound resource protection and improved off-policy importance sampling for agent training, using PyTorch and CUDA to optimize performance. He also addressed UI consistency by adding configurable action visibility. His work demonstrated depth in backend development, concurrency, and deep learning, delivering robust, maintainable solutions to complex engineering challenges.
Overall Statistics
Feature vs Bugs
75%Features
Repository Contributions
9Total
Bugs
2
Commits
9
Features
6
Lines of code
30,625
Activity Months4
Your Network
43 people
Work History
January 2026
1 Commits • 1 Features
Jan 1, 2026Concise monthly summary for 2026-01 focusing on Metta-AI/mettagrid. Delivered a kickstarting recipe for scripted agents, improved off-policy importance sampling calculations, and refreshed trainer defaults for CvC to boost performance and stability. No major bugs fixed this month; changes centered on enabling efficient CvC workflows and reproducibility.
1 Commits • 1 Features
Jan 1, 2026Concise monthly summary for 2026-01 focusing on Metta-AI/mettagrid. Delivered a kickstarting recipe for scripted agents, improved off-policy importance sampling calculations, and refreshed trainer defaults for CvC to boost performance and stability. No major bugs fixed this month; changes centered on enabling efficient CvC workflows and reproducibility.
January 2026
December 2025
1 Commits
Dec 1, 2025December 2025 monthly summary focusing on key accomplishments for Metta-AI/mettagrid. Delivered a targeted bug fix to improve UI clarity and control over actions. The Change Vibe Action Visibility bug was resolved by introducing a default-disabled configuration option, providing explicit control over its UI visibility and enhancing user experience. This work ties to issue #4262 and is documented in commit 00dbb1db8f5036a78acb36ac8f6e67670e8d90e5. Overall, the change improves UI predictability, reduces user confusion, and aligns visibility with access controls.
December 2025
1 Commits
Dec 1, 2025December 2025 monthly summary focusing on key accomplishments for Metta-AI/mettagrid. Delivered a targeted bug fix to improve UI clarity and control over actions. The Change Vibe Action Visibility bug was resolved by introducing a default-disabled configuration option, providing explicit control over its UI visibility and enhancing user experience. This work ties to issue #4262 and is documented in commit 00dbb1db8f5036a78acb36ac8f6e67670e8d90e5. Overall, the change improves UI predictability, reduces user confusion, and aligns visibility with access controls.
October 2025
3 Commits • 2 Features
Oct 1, 2025October 2025 for Metta-AI/metta: Delivered architectures that advance memory and sequence modeling capabilities, with clear business value in scalability, latency, and developer productivity. Key work includes Cortex, a modular memory architecture library with AxonCell/AxonLayer for meta-recurrence, GPU-accelerated via Triton kernels and integrated into the Metta framework; integration into mLSTM/sLSTM and the auto-stack builder (AMS); removal of a Triton stub and related compatibility fixes to stabilize production builds. Also delivered XLCell, a Transformer-XL style attention cell with stateful streaming, segment-aware rolling memory, and relative positional encoding; supporting optional AxonLayer-backed Q/K/V projections and integrated testing for both standard and AxonLayer-backed modes. These initiatives deliver faster inference, more flexible backbones for meta-recurrence, and improved developer workflows across the stack.
3 Commits • 2 Features
Oct 1, 2025October 2025 for Metta-AI/metta: Delivered architectures that advance memory and sequence modeling capabilities, with clear business value in scalability, latency, and developer productivity. Key work includes Cortex, a modular memory architecture library with AxonCell/AxonLayer for meta-recurrence, GPU-accelerated via Triton kernels and integrated into the Metta framework; integration into mLSTM/sLSTM and the auto-stack builder (AMS); removal of a Triton stub and related compatibility fixes to stabilize production builds. Also delivered XLCell, a Transformer-XL style attention cell with stateful streaming, segment-aware rolling memory, and relative positional encoding; supporting optional AxonLayer-backed Q/K/V projections and integrated testing for both standard and AxonLayer-backed modes. These initiatives deliver faster inference, more flexible backbones for meta-recurrence, and improved developer workflows across the stack.
October 2025
September 2025
4 Commits • 3 Features
Sep 1, 2025September 2025 monthly summary for Metta-AI development work focusing on key feature deliveries, bug fixes, and overall impact across two repositories (Metta-AI/metta and Metta-AI/mettagrid).
September 2025
4 Commits • 3 Features
Sep 1, 2025September 2025 monthly summary for Metta-AI development work focusing on key feature deliveries, bug fixes, and overall impact across two repositories (Metta-AI/metta and Metta-AI/mettagrid).
Activity
Loading activity data...
Quality Metrics
Correctness94.4%
Maintainability94.4%
Architecture94.4%
Performance88.8%
AI Usage31.2%
Skills & Technologies
Programming Languages
C++CUDACudaPyTorchPythonShell
Technical Skills
Attention MechanismsBackend DevelopmentBug FixingC++CI/CDCUDACode IntegrationCode IntrospectionCommand-line Interface (CLI)ConcurrencyData ProcessingDeep LearningGPU ComputingGame DevelopmentLibrary Design
Repositories Contributed To
Overview of all repositories you've contributed to across your timeline
Metta-AI/mettagrid
Sep 2025 – Jan 2026
3 Months active
Languages Used
C++Python
Technical Skills
Backend DevelopmentC++Game DevelopmentPythongame developmentConcurrency