I'm a Computer Science student pursuing an M.Sc. (Integrated) in Software Systems at PSG College of Technology.

I enjoy designing and building software that solves practical problems. My interests lie in software engineering, backend development, full-stack development, and algorithmic problem solving.

Currently, I'm focused on strengthening my programming fundamentals while building projects that help me understand how real-world applications are designed.

• Data Structures & Algorithms
• Backend Development
• System Design Fundamentals
• Artificial Intelligence & Machine Learning

Understand an unfamiliar codebase. Point CodeAtlas at a repository and it produces a browsable report computed from the code (Tree-sitter) and its git history (JGit), so every number traces back to a file, line or commit.

• Reading order — where to start reading, with files ranked by churn, reach and fan-in and the reasons behind each score
• Ownership — who last changed each line, bus factor per directory, single-owner and orphaned flags
• Change coupling & hotspots — files that change together, including hidden dependencies with no import between them, and code that is both busy and deeply nested
• Architecture — directories layered by import direction, with import cycles marked, plus an interactive import graph
• Findings — near-duplicate code, committed build output, untested areas and unimported modules, each with file-and-line evidence
• Ask — questions about the code; an optional LLM answers only from retrieved excerpts, and every citation is checked as exact, inside an excerpt, or unsupported

Tech: Java 17, Spring Boot, Tree-sitter, JGit, PostgreSQL, React, TypeScript, Docker.

API change-impact analysis. Diff tools say a change is breaking; Axon says who would break. It compares what an API promises (its OpenAPI spec), what it actually does (observed traffic) and what a proposed version would change.

• Impact analysis — breaking changes ranked by the clients that real traffic shows are affected; request-side impacts are observed, response-side ones only potential, and the two are never blended
• Drift detection — fields and operations that are documented but never used, or used but never documented
• CI gate — axon check fails a pull request when an observed breaking change hits a set share of active clients
• Privacy by design — captured traffic is reduced to field shapes in the browser; values never leave
• Web UI — clients × fields matrix, declared vs. observed contract, per-client timeline and an API map
• Evaluated in the open — the method was written before any result, and the write-up includes where Axon falls short

Tech: Java 17, Spring Boot, PostgreSQL, OpenAPI, React, TypeScript, Docker.

Multi-tenant Kanban project management: Trello-style boards with real access control and live collaboration. One deployment serves many organisations, with work organised into projects, boards and cards.

• Tenant isolation — enforced at the database layer; another organisation's data returns 404, so it is never confirmed to exist
• Two-layer roles — organisation (owner, admin, member) and project (head, member) roles, re-checked on every request so a demotion applies immediately
• Real-time boards — room-scoped Socket.IO updates; drag-and-drop with mouse or touch updates instantly and rolls back if the server rejects it
• Secure sessions — rotating, revocable, hashed refresh tokens; reusing an old token signs the user out everywhere
• Team features — email invites, @mention notifications, board search and filters, safe account deletion
• Tested — 761 tests, led by a dedicated cross-tenant isolation suite

Tech: TypeScript, React, Node.js, Express, MongoDB, Socket.IO.

NYC taxi demand forecasting and fleet repositioning. SurgeMap forecasts pickups in each of 258 taxi zones 5, 15, 30 and 60 minutes ahead, then tests in a simulator whether sending idle vehicles to where demand is coming shortens rider waits.

• Graph forecaster — a Graph WaveNet over the directed taxi-flow graph plus a zone graph it learns itself, trained on four months of NYC TLC trips
• Honest baselines — compared with persistence, historical average, ridge and gradient boosting on the same data, with bootstrap confidence intervals
• Fleet simulator — replays real pickups against a simulated fleet, with a min-cost-flow repositioning policy that can run on any forecast
• Interactive app — a Streamlit app to explore forecasts zone by zone and run the simulator
• Reproducible — one command downloads the data, a CLI runs every step, and runs are seeded

Tech: Python, PyTorch, NumPy, Pandas, Scikit-learn, SciPy, Streamlit.

Apart from coding, I enjoy exploring new technologies, learning through projects, and continuously improving my problem-solving skills.

I believe that building real-world applications is one of the best ways to grow as a software developer.

• 📧 Email: shrishaanth2024@gmail.com
• 💼 LinkedIn: www.linkedin.com/in/shrishaanth-dev
• 🌐 GitHub: https://github.com/shrishaanth

Feel free to reach out if you'd like to collaborate, discuss projects, or connect.