Compilers • Data/ML-ish(Calling an API lol)
Kasahara Lab @ Waseda University | Building real products + research-y systems
-success){kind=icon}
I'm a student who enjoys building things that are both practical and kinda technical.
- 🎓 BEng (Computer Science & Communications Engineering) @ Waseda University
- Kasahara Lab — Compilers & High-Performance Computing --> Working on Sub-Micro Clockgating for Greener Computing from the compiler level
- 🚀 Shipped a Flutter app to App Store + Google Play @ GomiMap
- 🪙 Doing Stablecoin and Market Maker Research @ TRADOM INC
- 📈 Built a PPO reinforcement learning agent for FX hedging (PyTorch + SB3) @ TRADOM INC
- 🔐 Little bit of experience with security reviews (AWS environment) @ TRADOM INC
- 🌏 Languages: English (native-level) • Japanese (JLPT N2) • Sinhala
- 🧵 Compiler / HPC performance experiments (Auto-Parallelizing Compiler + NAS Benchmarks + Clock-Gating for Energy Red)
- 📊 Market Making Research for Stablecoins (StableFX Sandbox + Info-Gathering)
- 🗺️ GomiMap: scaling waste-rule ingestion + data pipelines across cities + app dev
Cross-platform app shipped to App Store + Google Play, built with:
- Flutter + Firebase (Firestore DB) + Google Maps API
- Google Cloud for Data Cleaning/Processing workflows
- CI/CD via GitHub Actions
- City/ward-level waste rules → normalized schemas + ingestion pipeline + normalized dbs for schedule based on user area
🔗 Old Repo of Google Solution Challenge Submission: https://github.com/s3nmith/recycling-app
Current App Repo is Private :(
Web: https://gomi-map.jp/
Built a custom RL environment for hedging decisions:
- PPO agent with feature extractors for forward curves + exposure
- Backtesting + evaluation for stability & risk behaviour
- Reward decomposition (hedge error, costs, roll, etc.)
University research project benchmarking encrypted inference:
- CKKS approaches (e.g., TenSEAL/OpenFHE style workflows)
- Trade-offs across latency / accuracy / cost
- Compared against secure infrastructure options (e.g., confidential computing ideas)
Reproducible benchmarking harness:
- Makefile + Bash runner for NAS kernels (BT/CG/EP/FT/IS/LU/SP)
- Collected runtime + energy (RAPL via
perf stat) - CSV exports and scaling / Joules-per-op analysis
Reviewed AWS analytics environment for security improvement opportunities:
- SSH key management issues
- Local download & data egress risks
- Monitoring + activity log coverage gaps (CloudTrail/CloudWatch guardrails)
If you're building something in:
- compilers + benchmarking
- sustainability + real-world apps
- security stuff
…lmkk!