Software engineer in the making · Backend & systems · AI/LLM applications · Full-stack
CS undergrad at Scaler School of Technology · Bangalore, India
- 🎓 Computer Science undergrad at Scaler School of Technology, Bangalore.
- 🛠️ I like building things from the ground up: a database engine in C++, a sharded typeahead service, RAG pipelines and browser agents.
- 🚀 I ship: several of my projects are deployed, including a production website for an export business.
- 🔭 Currently deepening my skills in system design, distributed systems and LLM engineering.
- 🤝 Open to SDE internships: reach me at Samarthpd2112@gmail.com.
| Project | What it does | Stack |
|---|---|---|
| InsightLM · Live | NotebookLM-style document workspace: chat with PDFs and get grounded answers with page citations. Uses Corrective RAG with query rewriting, broad retrieval and judged re-ranking. | Next.js, Express, Qdrant, Qwen2.5, BGE embeddings, SSE |
| GradeSense | AI grading of student answer papers against rubrics, with evidence extraction, deterministic scoring, in-browser annotation and annotated-PDF export. | Next.js 16, TypeScript, Prisma, BullMQ, Redis, Gemini, PDF.js, Playwright |
| AgentFlow | Browser automation agent that reads screenshots with a vision model and decides how to click, type and navigate to finish tasks. | Python, Playwright, Gemini 2.0 Flash |
| WriteTex | Chrome extension that tailors LaTeX resumes and cover letters to a job inside Overleaf. Every edit is shown as a reviewable diff before it is applied. | TypeScript, Chrome Extensions API, LLMs |
| Outreach Engine | CLI pipeline for cold outreach that finds leads, researches companies, drafts evidence-backed emails, sends them via Gmail and tracks replies for A/B experiments. | Python, SQLite, OpenAI, Gmail API |
| Project | What it does | Stack |
|---|---|---|
| AxiomDB | Relational database engine built from scratch: buffer pool (LRU-K), B+Tree index, SQL parser, cost-based planner, Volcano executor, ACID transactions with strict 2PL, WAL crash recovery and a pluggable LSM-tree engine benchmarked against the B+Tree engine. | C++20, CMake, Catch2 |
| CoreTypeahead | Low-latency autocomplete service with a consistent-hash ring across Redis shards, time-decay trending scores and batched write-behind to SQLite (WAL). | Python, FastAPI, Redis, SQLite |
| RideShare · Live | Full-stack ride-hailing app with JWT auth, role-based access for riders and drivers, and full ride lifecycle tracking. | Spring Boot, MongoDB, Next.js |
| Multithreaded HTTP Server | HTTP/1.1 server built only on the standard library, with a thread pool, binary file transfer, JSON uploads and path-traversal and host-header protection. | Python (sockets, threading) |
| Rate Limiter | Thread-safe sliding-window rate limiter using the Strategy and Proxy patterns. | Java |
| Project | What it does | Stack |
|---|---|---|
| Veblyss Global | Production website for an Indian export business serving international customers across multiple product categories. | Next.js, TypeScript, Tailwind CSS |
More low-level design work: Parking Lot · Elevator · BookMyShow · Snakes & Ladders · ML research: Attention-head ablation on ViT/CIFAR-10
Languages: C++ · Java · Python · TypeScript · JavaScript · Kotlin · SQL
Backend: Spring Boot · Node.js / Express · FastAPI · REST · JWT · BullMQ
Frontend: React · Next.js · Tailwind CSS · Framer Motion
Data: MongoDB · MySQL · SQLite · Redis · Qdrant (vector DB) · Prisma
AI/ML: RAG pipelines · LLM agents · PyTorch · Hugging Face · Gemini / OpenAI APIs
Tools: Git · Linux · GitHub Actions · Playwright · CMake
I'm actively looking for software engineering internships. If you're hiring or want to build something together, I'd love to hear from you.
