SCROLL
SCROLL FOR WORK
Went deep into real ML engineering — building end-to-end pipelines (preprocessing, feature engineering, cross-validation, tackling class imbalance) and boosted-tree models like XGBoost. The part that lit me up most: LLMs — transformer internals, embeddings, and building RAG systems that ground generation in retrieved context with vector search.
On Team 3, I worked on lightweight deep-CNN architectures for medical imaging — profiling parameters, FLOPs, latency and accuracy, then applying pruning and compression to fit models onto FPGA and Android devices. An eight-week sprint from paper study to a full trade-off analysis, report and presentation.
RHCSA-level training on RHEL 9 — the real backbone of how servers actually run. Users & sudo delegation, LVM and storage, permissions and SELinux in enforcing mode, systemd, firewalld, dnf, and boot recovery via rd.break. This is where "it works on my machine" turned into "it works in production."
Client-side engineering done properly — semantic HTML5, modern CSS (flexbox, grid, responsive), ES6+ JavaScript, and React (components, hooks, virtual-DOM reconciliation). Learned to think about the browser's rendering pipeline, not just make things appear on screen.
A password manager unlocked by your signature — not a typed master password, but the way you sign. Shipped as a Chrome extension, it was my experiment in rethinking what "authentication" even has to look like.
An AI career mentor that turns "I don't know what to do next" into a real plan — tailored guidance, resources and direction for students figuring out their path.
Real-time emotion recognition from your webcam — face detection feeding a CNN that reads expressions live. My deep dive into computer vision, and I loved every debugging hour of it.
Control your cursor with nothing but hand gestures. Built with MediaPipe + OpenCV to see how far natural, touch-free interaction could go — turns out, pretty far.
A fair way to split fuel and trip costs, backed by smart contracts so nobody argues about the math. My first real crossover of app logic and blockchain — ambitious, and worth it.
A weather app with personality — animated day/night themes and a UI that reacts to the sky it's showing. Proof that even a "simple" idea deserves a delightful build.
An interactive birthday experience — a playful quiz that unfolds into personalised wishes. A reminder that code can be a gift, not just a tool.
A notebook-driven data & ML exploration for CEIP 2026 — where I get my hands dirty with real datasets, hunting for the signal in the noise.
My portfolio, containerized with Podman — because shipping it should be as clean as building it. My first proper step into DevOps and reproducible deploys.
My first portfolio site — where this whole journey of putting my work out into the world began. Rough around the edges, but honest, and I'm proud of the start.
The experimental engine behind my research paper — a harness that pits plain chunk-based RAG against curated, cross-linked "concept bundles," holding the corpus, generator model and token budget identical across both arms. Built with paired bootstrap significance testing and McNemar's test baked in, because a result you can't defend statistically isn't a result.
A free, ATS-aware resume scorer and builder — Next.js frontend, FastAPI backend, Postgres for storage, with a roadmap toward a distilled model doing the semantic scoring in later phases. My first real monorepo, and my first time architecting for phases I haven't built yet.

Winner — Smart India Hackathon 2025 (Internal Round). Led my team through a build-under-pressure sprint and came out on top. Nothing sharpens you like a deadline and a whiteboard full of ideas.

RHCSA certified — proving I can actually run and troubleshoot Linux systems in the real world, not just talk about them. My gateway into serious infrastructure.

Completed Red Hat's Enterprise Linux Automation with Ansible course; RHCE certification underway. Learning to make machines do the boring work so humans don't have to.

Google Cloud Data Analytics (Beginner) certificate. My first structured step into turning raw data into decisions — and I want to go much deeper.

Earned an Elite grade in IIT Madras' 12-week NPTEL Entrepreneurship course. Because building the tech is only half the story — I want to understand how ideas become ventures.

Built a working Li-Fi system with Arduino — transmitting data over light itself, using an LED to send and an LDR sensor to receive. Watching real text travel across a beam of light was the moment hardware truly clicked for me.
"Opportunities and Challenges for Improving Sustainable Transportation in India — E-Highway Buses" (STM Journals, Sep 2025). As sole author, I researched overhead-powered electric highway buses as a cleaner, longer-range alternative to diesel — infrastructure, cost, and renewable-energy feasibility for India. Proof that I don't just build systems, I can question and reimagine them.
Currently building the case that retrieval structure — not just quantity — is what curbs hallucination in RAG. Comparing naive chunk-based retrieval against curated, cross-linked "concept bundles" across single-hop, multi-hop and unanswerable questions on a 40-document corpus, with every answer hand-labelled by me before any number gets reported. Harness is live; numbers are still coming in.

Completion certificate for the Celebal Excellence Internship 2026, Data Science track (18 May – 18 Jul 2026, Enrollment ID CT_CSI_DS_251). This backs the Experience entry above with the actual paper.

Completion certificate for the LNMIIT Undergraduate Summer Internship Program 2026 — "Lightweight Models Development for Low-Power AI Applications," under Dr. Gaurav Chatterjee (01 Jun – 27 Jul 2026).

Completion certificate for the Frontend Development internship at GenX Soft Technologies (08–22 Jul 2024, Intern ID GSIN05181007).
Everything I've built and learned so far, in one page — and I'm just getting started. Grab a copy.
I'm a student, and I think that's my biggest advantage — I get to be curious full-time. I like problems that sit at the intersection of scale and simplicity, where the right abstraction makes a whole system quieter. Every project I take on is a bet that I'll come out the other side knowing something I didn't.
I don't claim to know everything — I claim to find out. Give me a problem I've never seen, and watch how fast that changes. Outside the editor I'm reading about systems design, playing chess, and chasing the small satisfaction of a green CI run.
Other platforms, communities and profiles I'm active on.
Deployed projects and published work — live and clickable.