Backend-focused software developer working with Python, Java, SQL, and web APIs. I convert notebooks and experiments into production-grade, tested, deployable systems.
I am pursuing B.Tech Information Technology at MAIT and the BS in Data Science and Applications at IIT Madras. I like building practical backend systems, ML pipelines, developer tools, and data-driven applications that are easy to run, test, and deploy.
Coursework and IIT Madras project repositories are on my education account: github.com/23f3001694
- Backend Developer Intern experience with Django REST Framework and API migration work
- Building ML projects as production pipelines (config-driven, tested, Dockerized, CI/CD)
- Strengthening Java backend skills with Spring Boot, JPA, PostgreSQL, and JUnit
- Practicing DSA and problem solving for entry-level software engineering roles
| Project | Stack | Highlights |
|---|---|---|
| BehaveGuard | FastAPI, PyTorch, Next.js, scikit-learn, LSTM, TCN | Continuous biometric authentication (keystroke rhythm + mouse kinematics) using multi-model score fusion (30% SVM + 35% LSTM + 35% TCN), speed-scale invariance, and real-time sandbox test environment |
| Comment Toxicity | Python, LightGBM, scikit-learn, Docker, GH Actions | 4-class classifier with 66+ features, OOF target encoding, Nelder-Mead blend optimization, per-class threshold tuning — 0.817 macro F1 on 198K samples |
| LeafSR | PyTorch, GAN, Docker, GH Actions | ESRGAN-lite super-resolution pipeline for plant leaf images — Charbonnier + perceptual + FFT losses, EMA checkpointing, 8-way TTA, 7.2% MAE reduction over bicubic baseline |
| Project | Stack | What it shows |
|---|---|---|
| Cortex | Vue 3, Flask, Celery, Redis, ChromaDB, Gemini, Gemma | RAG codebase Q&A companion, dynamic repo indexing (AST-aware function/class chunking), PR & diff integration, Server-Sent Events (SSE) typewriter-style streaming |
| Keyway | Flask, Redis, Vue 3, PostgreSQL, Docker, PBKDF2 | Self-hosted secrets vault with a zero-persistence model for raw decryption keys (PBKDF2-HMAC-SHA256 with 480k iterations), session-only Redis/RAM storage, outbound key health checks, and secure redirection/method constraints |
| Stash | Flask, SQLite, SQLAlchemy, Docker, Vanilla JS | Lightweight URL stashing utility prepended via address bar. Features trailing-slash insensitive deduplication, interactive monospace UI with checkbox status, and standalone portable HTML exports |
| Telegram Job Scraper | Python, Gemini AI, Telegram Bot API, Docker | Lightweight scraper filtering job postings via Gemini AI, deduplicating records, and broadcasting alerts to Telegram channels |
| TDS Quiz Solver | FastAPI, LangGraph, Playwright, Docker | LLM agent orchestration, sandboxed code execution, browser automation, API design |
| Clip Rewrite | FastAPI, SQLAlchemy, Jinja2, Docker | Temporary text/file sharing, uploads, syntax highlighting, expiration workflows |
Languages: Python, Java, SQL, Bash, JavaScript, HTML/CSS
ML & Data: PyTorch, scikit-learn, TensorFlow (basic), LightGBM, Pandas, NumPy, Jupyter, ChromaDB
Backend: FastAPI, Flask, Django REST Framework, SQLAlchemy, REST APIs, Celery, Redis, Pytest
Databases: PostgreSQL, MySQL, SQLite, MariaDB, MongoDB
DevOps & Tools: Git, Docker, AWS (EC2, S3), Linux, GitHub Actions (CI/CD), Postman, macOS
- Portfolio: amehta.space
- Resume: briskam.github.io/resume
- github.com/BriskAM · github.com/23f3001694
- LinkedIn: akshit-mehta2005
- Email: akshit.mehta.work@gmail.com