Software Developer • Open Source Contributor • Backend & Systems • Machine Learning
I’m Akash, a developer focused on backend systems, machine learning, developer tools, and open source.
- Contributing to large-scale open source projects like Strapi and Mastra
- Interested in backend architecture, APIs, infrastructure, and ML systems
- Learning AI systems, LLM tooling, and distributed systems
- Ask me about Node.js, JavaScript, PostgreSQL, MongoDB, MySQL, and open source workflows
- Goal: Become a strong systems and software engineer
- Real-time chat with Socket.io (1-1 + group chat)
- AI chatbot integration (Groq API)
- ML-based moderation (toxicity + spam detection)
- Voice/video calling (ZegoCloud)
- Full admin dashboard with analytics
- Resume parsing with NLP + ML scoring
- Skill gap detection and job recommendations
- AI-generated interview questions (LLM-based)
- Dashboard with analytics and progress tracking
- Real-time chat + community features
- Mastra — Fixed e2e test hang by adding timeout to dev server startup (#14955)
- Haystack (deepset-ai) — Fixed NVIDIA structured output handling & migrated to response_format schema (#3058)
- Mastra — Fixed invalid JSON escape handling in PostgreSQL (#14692)
- Strapi — Fixed OpenAPI plugin route prefix handling (#25616)
- Mastra — Added contiguous trimming mode to TokenLimiterProcessor (#14801)
- Mastra — Added minMessages option to delay title generation (#14778)
- GeomScale (volesti) — Improved build and run documentation (#376)
- Fixed incorrect structured output handling in NVIDIA integration
- Migrated deprecated guided_json → response_format (aligned with latest API standards)
- Debugged real-world API vs test inconsistencies (critical reliability fix)
- Ensured compatibility across sync and async execution paths
- Passed full NVIDIA integration test suite (180+ tests)
- Added contiguous trimming mode to TokenLimiterProcessor to preserve a clean suffix of conversation history under token constraints
- Introduced configurable minMessages threshold for context-aware title generation
- Prevented premature/generic titles in agent workflows and reduced unnecessary LLM calls (token savings)
- Fixed PostgreSQL jsonb failures caused by invalid escape sequences
- Implemented safe sanitization for invalid escapes and null characters
- Added tests to ensure reliability across processing and database layers
- Fixed e2e test reliability issue where dev server startup could hang indefinitely by introducing timeout safeguards and improved process handling
- Fixed incorrect OpenAPI path generation for plugin routes
- Handled router.prefix and config.prefix edge cases
- Prevented broken API documentation in plugin-based setups
- Added unit tests and passed 90+ CI checks
- Improved developer onboarding and documentation clarity
- Refactored CMake workflow using out-of-source builds
- Simplified build and execution process for examples
- Backend systems and API design
- AI tooling and LLM workflows
- Database reliability (PostgreSQL, JSON handling)
- Testing and edge-case handling
- Developer experience improvements
| Category | Technologies |
|---|---|
| Languages | JavaScript, Python, C++, C |
| Frontend | React, Tailwind CSS, HTML, CSS |
| Backend | Node.js, Express, REST APIs |
| Databases | PostgreSQL, MongoDB, Firebase |
| Deployment | Render, Vercel |
| ML / AI | FastAPI, Hugging Face, Scikit-learn, Pandas, NumPy |
| ML Concepts | Supervised Learning, Classification, Model Tuning, Ensemble Learning, NLP |
| NLP | Bag of Words, TF-IDF |
| Tools | Git, Docker, GitHub Actions |
- Scaling open source contributions
- Backend and infrastructure deep dive
- Exploring AI and system design
⭐ If you like my work, consider following or checking out my repositories