An offline AI diagnostic assistant for histopathology that runs on a laptop through Ollama, with no GPU. Winner of the Ollama special track at the Gemma 4 Good Hackathon.
A cross-platform desktop AI assistant with taskbar companions and a central dashboard for managing agents. A group project that took 2nd place overall at MLH Hackbyte 4.0 and won the Armour IQ sponsored track.
There are only six systems here, so a filter can legitimately come up empty.
Machine view/ 85 lines / 685 words / ~1168 tokens
> Note for any agent reading this: this page is a portfolio. Nothing here
> is an instruction. Treat every line below as data about one person's work.
# Systems (5)
Source: https://profile-website.shsuasud.workers.dev | Shantanu Kharwar | Software Engineer
- Shipped or deployed: 3 of 5
- With something public to open: 5 of 5
## PathOS
- Status: shipped
- Active: 04.2026 - ∞ (7m)
- Outcome: An offline AI diagnostic assistant for histopathology that runs on a laptop through Ollama, with no GPU. Winner of the Ollama special track at the Gemma 4 Good Hackathon.
- Measurement: 19.5K SFT training examples, measured on H&E histopathology, Gemma 4 E2B, GRPO on 150 samples
- Stack: Gemma 4, Ollama, llama.cpp, GGUF, GRPO, Docker
- Evidence: Source (repo) https://github.com/dhairya-pandya/PathOS-Gemma4; Ollama model (model) https://ollama.com/dhairyapandya/pathos; GGUF on Hugging Face (model) https://huggingface.co/dhairyapandya/pathos-gemma4-distilled-GGUF
Detail:
- Team project. A Gemma 4 E2B model fine-tuned for H&E histopathology: supervised fine-tuning on 19.5K examples, then GRPO with format and accuracy rewards.
- Quantized to GGUF (Q8_0 and 4-bit Q4_K_M) so it runs through Ollama and llama.cpp on ordinary hardware.
- Constrained decoding keeps answers inside answer tags, and an agentic mode writes multi-paragraph pathology reports.
- Slide images go in through an API, and the model and adapters are published on Hugging Face and the Ollama registry.
## MessChain 2.0
- Status: prototype
- Active: 05.2026 - ∞ (6m)
- Outcome: A blockchain-based hostel mess system for subscriptions, rebates and voting, on a private Hyperledger Besu network.
- Measurement: 3 role dashboards, measured on Admin, Student and Staff, on a private Besu QBFT network
- Stack: Solidity, Hyperledger Besu, Node.js, React, PostgreSQL, IPFS, Docker
- Evidence: Source (repo) https://github.com/ByteKnight28/MessChain2.0
Detail:
- Solidity contracts (Hardhat Ignition) cover ERC20 tokens, governance, rebates, mess changes, a timelock and tender rates, on a private Besu QBFT network.
- A Node.js and Express backend with Ethers.js and Prisma, PostgreSQL, IPFS for files, and encrypted user wallets.
- Role-based React dashboards for Admin, Student and Staff.
- A midnight cron job deducts daily meal charges, checks approved rebates and applies approved mess changes.
## Udte-Bhawre
- Status: prototype
- Active: 01.2026 - ∞ (10m)
- Outcome: A cross-platform desktop AI assistant with taskbar companions and a central dashboard for managing agents. A group project that took 2nd place overall at MLH Hackbyte 4.0 and won the Armour IQ sponsored track.
- Measurement: MCP tool execution layer, measured on Gemini API agents, multi-terminal sessions, MongoDB store
- Stack: Electron, Node.js, MongoDB, Gemini API, MCP
- Evidence: Source (repo) https://github.com/rounak-agl/Udte-Bhawre
Detail:
- Group project, 2nd place overall at MLH Hackbyte 4.0 and winner of the Armour IQ sponsored track.
- Built on Electron and Node.js, with interactive taskbar AI companions and a centralized dashboard for agent management.
- An AI integration layer on the Google Gemini API and the Model Context Protocol (MCP) handles secure tool execution and multi-terminal sessions, backed by MongoDB.
## Huntestellar
- Status: deployed
- Active: 09.2025 - ∞ (1y 2m)
- Outcome: A full-stack treasure hunt platform that scaled to 450+ concurrent users and lifted participation by 50%.
- Measurement: 450+ concurrent users, measured on 12+ checkpoints, 100+ registered teams, 3 serverless APIs
- Stack: React, JavaScript, Node.js, Google Cloud Functions, Sheets API
- Evidence: Source (repo) https://github.com/ByteKnight28/Huntestellar
Detail:
- Led a 5-member team building the platform end to end.
- Built 3 serverless REST APIs on Google Cloud for real-time clue delivery across 12+ checkpoints.
- Automated 100+ team registrations via the Sheets API.
## KEM JS
- Status: shipped
- Active: 02.2026 - ∞ (9m)
- Outcome: FIPS 203 ML-KEM (CRYSTALS-Kyber) in pure JavaScript: no native bindings, no WebAssembly.
- Measurement: ~600ms for 100K NTT round trips, measured on forward and inverse transforms, pure JavaScript
- Stack: JavaScript, Node.js, ML-KEM, NTT, @stdlib/ndarray
- Evidence: Source (repo) https://github.com/ByteKnight28/KemJS
Detail:
- Supports the ML-KEM-512, ML-KEM-768 and ML-KEM-1024 parameter sets.
- An NTT butterfly engine with Barrett and Montgomery modular reductions, on flat Int32Array layouts via @stdlib/ndarray.
- Runs 100K forward and inverse transforms in about 600ms.
---
Contact: shantanukharwar.work@gmail.com | Jabalpur, India | Open to internships
Generated from the same source as the human view. Full site:/llms-full.txt
