Skip to content
View harsh-pandhe's full-sized avatar
🛰️
🛰️ Prepping IRoC-U 2026 Finals
🛰️
🛰️ Prepping IRoC-U 2026 Finals

Block or report harsh-pandhe

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
harsh-pandhe/README.md

Hi there, I'm Harsh Pandhe

Harsh Pandhe is a robotics and systems engineer based in Navi Mumbai, India. He builds autonomous navigation stacks (ROS 2, PX4, LiDAR SLAM), multi-agent reinforcement learning for robot swarms, quantitative trading research, and geospatial infrastructure software. He is a Software Development Intern at Doppelmayr, a Finalist and Automation & Navigation Lead at the ISRO Robotics Challenge (IRoC-U 2026), the Smart India Hackathon 2025 National Winner, and an SSRN-published quantitative finance researcher.

Teensy firmware, a Rust order-matching engine, Sentinel-2 pipelines, ROS 2 swarms, Solidity contracts. Ninety-one public repositories in, the range is less scattered than it looks — I keep landing in the same place: systems that touch the physical world, where being wrong is expensive and nobody grades you on the demo.

Portfolio: harshpandhe.com LinkedIn profile of Harsh Pandhe SSRN published paper by Harsh Pandhe Dev.to blog Email Harsh Pandhe

Roles: IRoC-U 2026 Finalist, SIH 2025 National Winner, SSRN-published quant researcher, SDE Intern at Doppelmayr


Quick facts

name:        Harsh Pandhe
pronouns:    he/him
role:        Software Development Intern, Doppelmayr (Apr 2026 – present)
focus:       Autonomous navigation · Multi-agent robotics · Quantitative research · Geospatial infrastructure
location:    Navi Mumbai, Maharashtra, India
education:   B.Tech Computer Science, Sinhgad Institute of Technology (SIT) Lonavala, 2024–2027
awards:
  - Smart India Hackathon 2025 — National Winner (team lead, 10,000+ teams)
  - ISRO Robotics Challenge IRoC-U 2026 — Finalist, Automation & Navigation Lead (Team ID 10144)
  - RoboRashtra 2026 — 1st Runner-Up, Chakravyuh maze & line solving
publications:
  - SSRN (2026) — "When Gradient Boosting Does Not Beat Momentum" (quantitative finance)
  - IRJMETS (2024) — "Collaborative Real-Time Code" (distributed systems)
leadership:  Web, Cybersecurity & Blockchain Head, IIC-SIT · Lead Organizer, Project Morpheus 2026
core_stack:  [ROS 2, PX4, Gazebo, Python, PyTorch, Rust, TypeScript, Next.js, PostgreSQL, Solidity, Docker]
open_source: 91 public repositories · upstream contributor to validator.js and Textualize/rich
contact:     harshpandhehome@gmail.com · https://harshpandhe.com · https://github.com/harsh-pandhe
open_to:     Research collaborations · Open-source work · Robotics & quant roles

Navigate: Currently · Research · Robotics · Quant · Geospatial · Full-stack & AI · How I work · Experience · Stack · All projects · Contact


🟢 What I'm working on right now

  • 🛰️ ISRO Robotics Challenge (IRoC-U 2026) — Finalist and Automation & Navigation Lead. GPS-denied UAV navigation: Fast-LIO over a Livox MID-360 LiDAR, EKF2 fusion with optical flow, ROS 2 + PX4, a Raspberry Pi 5 companion computer split from a Pixhawk 6c flight controller, and a custom docking station for persistent missions.
  • 💼 Doppelmayr — six months owning navigation and automation work on DoppelDash and ropeway automation, inside confidential projects valued in the millions, against strict international engineering standards.
  • 🎆 Taramandal — a full drone-show stack: taramandal-studio (parametric swarm choreography, bezier-spline pathing, music timeline sync, OBJ voxelizer, cKDTree collision validation) exporting straight into Taramandal-GCS (PX4 + Gazebo SITL, MAVSDK bridge, 5 Hz WebSocket telemetry, auto-abort failsafes).
  • 📊 AlgoTrade — a live paper-trading system on the NSE, now on a bounded 60-day validated run with automatic proof reporting. Two overlays promoted this quarter; four rejected on the data.
  • ⏱️ ChronoTrack v3.2.5 — 2FA (TOTP + WebAuthn passkeys), team-lead dashboards with productivity heatmaps, searchable audit trail, electron-updater auto-update.

↻ Updated 2026-08-02


📄 Research & publications

When Gradient Boosting Does Not Beat Momentum

A Significance-Tested, Walk-Forward Negative Result on Cross-Sectional Stock Selection in the Nifty 500 · SSRN, 2026 · independent researcher

🔗 Read the paper on SSRN

An XGBoost cross-sectional ranker is tested head-to-head against a plain 12-1 momentum baseline on the Nifty 500, under a strict walk-forward protocol with realistic transaction costs. The result is negative and reported as such: the machine-learning model does not beat the baseline once statistical significance and costs are applied. Even a native learning-to-rank objective that matches the baseline's Sharpe ratio needs 1.8× the portfolio turnover to get there — the added complexity buys no distinguishable benefit.

Evaluation used the Stationary Block Bootstrap and Jobson–Korkie–Memmel tests. The open-source research harness is published alongside the paper.

Collaborative Real-Time Code

IRJMETS, January 2024 — collaborative editing architecture; 50+ concurrent users at sub-50 ms sync (implemented as CollabCodex).


🤖 Robotics & autonomy

Cooperative area coverage with a TurtleBot3 swarm in ROS 2 Jazzy + Gazebo.

  • MAPPO / CTDE with a centralized critic for multi-agent credit assignment
  • A permutation-invariant GNN policy that scales to any fleet size without retraining — neighbour embeddings pooled through a shared MLP, invalid agents masked
  • A Control Barrier Function safety filter solving a QP at 20 Hz to guarantee zero collisions, with nominal-control fallback so a failed solve can't deadlock the robot
  • Active Collision Avoidance Safeguard, range-limited consensus map sync, semantic vision, 100 % area coverage in continuous-exploration evaluation

Python · ROS 2 Jazzy · Ray RLlib · PyTorch · PettingZoo · Gazebo

Ground-control and Software-in-the-Loop suite for autonomous drone-show fleets.

  • Hungarian-assignment formation morphing between choreography states
  • Layered safety loop: pre-flight launch-geometry check (0.5 m tolerance), live proximity rings (2.0 m lateral separation → immediate RTL), a 30 m cylinder geofence triggering emergency landing
  • FastAPI + MAVSDK gRPC-to-REST bridge streaming telemetry at 5 Hz over WebSockets; React Three Fiber 3-D previewer; .skyc import; RTK and latency telemetry
  • Companion taramandal-studio generates and safety-validates shows, then hands off to the GCS in one click

Python · PX4 · Gazebo · MAVSDK · FastAPI · React · Three.js

RoboRashtra 2026 — 1st Runner-Up (₹10,000). A hybrid bot that switches from high-speed deterministic line-following to neural-network inference the moment the maze begins, with a safety supervisor override on top. Spatial distance data → motor commands through hidden layers, running on-device; motor encoders mapped pulse-to-distance so a 90° turn is exactly 90°.

C++ · Teensy 4.1 · Neural networks · N20 encoders


📊 Quant & systems

AlgoTrade — systematic NSE paper-trading system private

A live, 24/7 momentum strategy on the Nifty 500 with the operational scaffolding to match: HMM regime detection, XGBoost experiments, cross-sectional momentum, a local Ollama LLM layer, Telegram alerting, SEBI 2026 compliance, SQLite WAL persistence, automated backup rotation.

What it's actually taught me is failure modes. A partial data-vendor outage returned NaN prices rather than missing ones; price or entry_price let them through, because float('nan') is truthy in Python. Downstream, abs(nan) > 1.0 evaluated to False, so the drift monitor reported "all clear" on top of a corrupted NAV for two days — a fully green dashboard sitting on bad data. Fixed in three layers (filter at source, explicit math.isnan(), monitor treats NaN as hard failure) plus seven regression tests.

The research arm of this system produced the SSRN paper above.

Python · XGBoost · hmmlearn · stable-baselines3 · Upstox V3 · Ollama · Flask

Rust order-matching engine

A limit-order-book matching engine written in Rust — price-time priority, the parts of exchange infrastructure where correctness under concurrency is the whole problem.


🛰️ Geospatial & civil infrastructure

terra-verify — construction progress from satellite imagery private

Infrastructure intelligence from Sentinel-2: automated measurement of construction progress from multispectral satellite imagery. Train/test separation is enforced in CI, not in good intentions — the pipeline fails the build rather than trusting a developer to remember.

Python · Sentinel-2 · Rasterio · scikit-learn

RopeIQ / Autonomous-Ropeway-Alignment — Eurocode-validated optimizer private

DEM + two station coordinates → permit-grade aerial ropeway alignment (towers, catenary, safety envelope, DXF, LandXML, Google Earth KML, 3-D twin, BOM, capex) in under 15 seconds on a laptop.

  • Validated against 19 real installations across 6 system archetypes on 5 continents — Aiguille du Midi 0.6 %, Zugspitze 2.7 %, Whistler Peak2Peak 7.2 %
  • Catenary mechanics validated to 0.00 % against three textbook references (sag / length / tension)
  • Eurocode envelope: EN 12929-1 clearance · EN 1991-1-4 wind · ISO 12494 ice · EN 1993-1-1 tower buckling · ANSI B77.1 live load
  • Four optimizers benchmarked head-to-head: DEAP GA · pymoo NSGA-II · stable-baselines3 PPO · scikit-learn RSM surrogate
  • REST service with JWT auth, RBAC, platform-owner role separation, audit log, immutable project snapshots; 157 pytest cases, CI on every push

Python · FastAPI · PostgreSQL · PyVista · Docker · Copernicus GLO-30 DEM

linear-scheduler-app — Line-of-Balance scheduler private

Linear scheduling (LoB / LSM) for heavy civil infrastructure — highways, railways, pipelines, ropeways. Where a Gantt chart shows only time, this plots time against chainage. Phase 15 added a resource-leveling engine, print composer, and DXF/raster backgrounds.

pod-transit — personal rapid transit digital twin private

Browser-based simulator for on-demand pod-transit (ATN) networks: synthetic-population OD demand, policy-gradient dispatch, infra cost/ROI modelling against rail and BRT, SUMO/MATSim interop, 481 tests. Era 2 adds vendor-agnostic vehicle profiles, Right-of-Way exclusion zones with cost penalties, KML/KMZ + GeoJSON round-trip, and an Expo SDK 56 rider app.


🌐 Full-stack, AI & Web3

Project What it is Stack
AtherCopilot RAG workspace for real-time PDF & URL analysis — embedding store, streaming responses, Clerk → Firebase token-minting bridge for auth-to-DB identity propagation Next.js 15 · Gemini 2.5 Flash · Google Genkit · Firebase
airgapped-netops-ai NOC copilot that predicts network degradation and explains itself — zero outbound traffic. Citation-aware RAG, Isolation Forest anomaly detection, SHAP explainability, retraining feedback loop FastAPI · React · Ollama · ChromaDB
Project Morpheus 2026 Logistics engine + site for a national 24-hour hackathon — 150+ teams, 5 domains, ₹2,00,000+ in prizes. Rule-driven allocation, atomic MongoDB transactions preventing double-booking under concurrent load. Live Next.js · MongoDB · Three.js · Framer Motion
CityConnect Civic reporting with on-chain EcoCoin rewards — 1000+ req/s with Redis caching Next.js · Django · Solidity · MongoDB
Telhan Sathi AI oilseed-farming advisory — SIH 2025 National Winner (MoA&FW PS 25273). ML crop-shift models, IoT soil sensing, PWA with real-time farm dashboards Python · Django · IoT · PWA
AgriTrace Circular-economy platform connecting farmers to recycling agents — AI waste classification + Razorpay traceability ledger Next.js · Google GenAI · Razorpay · Firebase

🧭 How I work

I would rather publish a negative result than a flattering backtest.

My trading research has killed four hypotheses for every two it promoted, and the one paper I've published to SSRN reports a model that didn't work. My satellite pipeline enforces train/test separation in CI rather than in good intentions. My drone GCS refuses takeoff over a 0.5 m placement error.

A recent example: mid-project on a helmet-detection model sitting at 368 training images, the obvious move was a 91,000-frame, 29 GB dataset. Instead I retrained on 25/50/75/100 % of what I already had and plotted the curve — 0.49 → 0.62 → 0.70 → 0.72 mAP@50, with the last quarter of data buying +0.02. The model wasn't data-limited. I skipped the download and fixed label quality instead. The technique is page one of any ML course; what's rare is running it before scaling the dataset rather than after.

This is the same instinct across every domain I work in: measure the thing before you buy it, and let the measurement be allowed to say no.


💼 Experience & leadership

Software Development Intern — Doppelmayr

Apr 2026 – present · ropeway & transport automation, global engineering team

Building DoppelDash, an internal operational platform (TypeScript, React, Node.js, MongoDB) — role-aware global search, approval chains, audit logging with plain-English summaries, IMAP inbox integration, branded PDF/Excel exports, and a full-application security-hardening pass. Contributing to confidential projects valued in the millions, where every line of code must meet strict international standards. Equipped with an RTX 4060 workstation for physics simulation, LLM workflows, and enterprise AI automation. 📰 Featured in Maval Maza — state-level press coverage of the selection.

Automation & Navigation Lead — ISRO Robotics Challenge (IRoC-U 2026)

Finalist, Team ID 10144 · official shortlist

Leading GPS-denied autonomous UAV navigation: Fast-LIO SLAM over Livox MID-360, EKF2 sensor fusion with optical flow, ROS 2 + Nav2 + PX4 SITL, Pi 5 / Pixhawk 6c companion-compute split over MAVLink, Coral TPU thermal + RGB fusion for search-and-rescue, and a custom docking station with wireless charging for persistent missions.

Team Member(Unofficial Lead) — Smart India Hackathon 2025 (National Winner)

Team Algo Sapiens · Dec 2025

Led six people to national victory among 10,000+ teams with Telhan Sathi — the first team from SIT Lonavala ever to reach nationals, and to win. Architected the ML crop-shift models and IoT soil-detection hardware, interfaced the hardware to the app live when it failed during the 36-hour finale, and scripted the pitch sequence itself.

Web, Cybersecurity & Blockchain Head — IIC-SIT

Oct 2025 – present · Institution's Innovation Council, SIT Lonavala

Overseeing web, security and blockchain across campus initiatives for 500+ students. Lead Organizer of Project Morpheus 2026 — a national 24-hour hackathon with 150+ teams across 5 domains and ₹2,00,000+ in prizes. Co-architected Kalpataru 2025. Mentored 70+ developers through the CODE2LAUNCH workshop; evaluated 50+ projects on architecture and security.


🛠️ Technical stack

Robotics & autonomy — ROS 2 (Jazzy) · Nav2 · PX4 SITL · MAVSDK · MAVLink · Gazebo · Isaac Sim · RViz · Pixhawk 6c · Livox MID-360 LiDAR · Fast-LIO · EKF2 · Raspberry Pi 5 · Coral TPU · Teensy 4.1 · motor encoders · PID control

Machine learning — PyTorch · Ray RLlib · PettingZoo · MAPPO/CTDE · graph neural networks · Control Barrier Functions · XGBoost · hmmlearn · stable-baselines3 · scikit-learn · SHAP · TensorFlow · Google Genkit · Gemini 2.5 Flash · Ollama · RAG / ChromaDB

Languages — Python · TypeScript · JavaScript · Rust · C++ · Solidity · Java · SQL · Bash

Backend & data — FastAPI · Node.js · Express · Django · PostgreSQL · PostGIS · MongoDB (transactions) · Redis · SQLite WAL · Prisma · Firebase · WebSockets · gRPC · Celery · Ray

Frontend — Next.js 15 · React · React Three Fiber · Three.js · Electron · Expo / React Native · Tailwind · Framer Motion

Geospatial — Sentinel-2 · Copernicus GLO-30 DEM · SRTM · PostGIS · Rasterio · PyVista · CesiumJS · LandXML · DXF · KML/KMZ · GeoJSON · SUMO / MATSim

Blockchain & Web3 — Solidity · Ethereum · smart contracts · Hyperledger Fabric · IPFS

Infrastructure — Docker · GitHub Actions · AWS · GCP · Vercel · Linux · pytest · Playwright


📚 Full project index

All 91 public repositories — expand for the full list
Project What it is Stack Status
🎆 taramandal-studio Swarm choreography generator & 3-D visualizer — parametric show design, bezier-spline pathing, music timeline sync, OBJ voxelizer, cKDTree safety validator, one-click Send-to-GCS JavaScript · React · Three.js Active
🛰️ Taramandal-GCS Drone-swarm command & control + SITL suite — Hungarian formation morphing, proximity rings, geofence auto-abort, RTK telemetry, .skyc import Python · PX4 · FastAPI · React Active
🤖 MARS Multi-agent TurtleBot3 swarm — MAPPO/CTDE, permutation-invariant GNN policy, 20 Hz CBF safety QP, 100 % coverage Python · ROS 2 · Ray RLlib · PyTorch Phase 2
🤖 Bot Teensy 4.1 neural maze solver + line follower — RoboRashtra 2026 1st Runner-Up C++ · Teensy 4.1 · Neural networks Competition winner
🌐 airgapped-netops-ai Fully offline NOC copilot — citation-aware RAG, Isolation Forest anomaly detection, SHAP explainability FastAPI · React · Ollama · ChromaDB Tracks 1–3 complete
⏱️ ChronoTrack Privacy-respecting workforce telemetry & project-ROI platform — Electron agent + Python AI daemon + web dashboard, GDPR/DPDP-compliant, 2FA with TOTP & passkeys Electron · Python · React · PostgreSQL v3.2.5
🤖 AtherCopilot RAG AI workspace — multi-document context, Clerk → Firebase auth bridge Next.js 15 · Genkit · Gemini · Firebase Live
🎯 Project-Morpheus National 24-hour hackathon platform — 150+ teams, atomic allocation engine Next.js · MongoDB · Three.js Live
🌱 AgriTrace Agricultural waste traceability — AI classification + Razorpay ledger Next.js · Google GenAI · Firebase Active
🎮 3d-chess 3-D multiplayer chess — Stockfish AI, Supabase realtime, glassmorphic UI Next.js · React Three Fiber · Supabase Live
🚁 Drone Scrollytelling drone-workshop site — 40-frame scroll sequence, isometric LiDAR swarm sim on Canvas Next.js 14 · TypeScript · Canvas API Live
🚁 Fly ASCEND — aerial intelligence landing page (autonomous flight, precision data capture) HTML · JS · 3D Live
❄️ SmartCooler-IoT ESP32 cooling controller — relay + DHT22 + DS18B20, humidity auto-mode, NVS persistence, web dashboard ESP32 · C++ · Firebase Deployed
😴 Drowsiness-Detection-System Real-time eye-aspect-ratio drowsiness detector for driver safety MediaPipe · Node.js · OpenCV Built
📅 365DaysOfCodeChallenge2k25 A year-long daily coding challenge, in public Python · Web Ongoing
📡 Lidar · thermal · UWB Sensor-integration research modules for the ASCEND / IRoC drone stack Python Research
🛰️ iroc-ascend-nav private IRoC drone nav stack — simulation-first on a 1-year pure-sim roadmap; pure-Python 3-D occupancy mapping + 3-D A* planning, Three.js dashboard, Mars heightmap terrain Python · ROS 2 · NumPy · Three.js Simulation-first
📊 AlgoTrade private Systematic NSE momentum system — HMM regime, XGBoost research, Ollama LLM layer, Telegram ops, SEBI 2026 compliance; source of the SSRN paper Python · Upstox V3 · XGBoost · Flask Runs 24/7
🚡 RopeIQ / Autonomous-Ropeway-Alignment private Eurocode-validated ropeway alignment optimizer — 19 real installations, four optimizers, DXF/LandXML/KML export Python · FastAPI · PostgreSQL · PyVista Phase 1 product
🛰️ terra-verify private Construction progress measured from Sentinel-2 imagery; train/test separation enforced in CI Python · Sentinel-2 · scikit-learn Active
🚉 pod-transit private PRT/ATN digital-twin simulator — OD demand, RL dispatch, cost/ROI vs rail & BRT, SUMO/MATSim interop, 481 tests React · TypeScript · Three.js · Expo Era 2
🛣️ linear-scheduler-app private Line-of-Balance scheduler for highways, railways & pipelines — resource leveling, print composer, DXF backgrounds Next.js · PostgreSQL · Prisma Phase 15
🛠️ GripGroup private Generative geospatial design for linear infrastructure — SRTM/PostGIS, multi-objective optimizer, LandXML export, Ray + Celery, credit-ledger billing Python · React · PostGIS · Ray Phase 9
🌾 KrishiPancham-AI private Agricultural AI companion — Panchang almanac + weather forecasts + NDVI crop health + 32 govt schemes; multilingual, voice I/O, offline-first FastAPI · Expo · Next.js · ephem Phases 0–5
📱 ExpenseTracker private Production React Native app — bank SMS parser for 10+ Indian banks, OTA updates, 4-layer dedup sync React Native · TypeScript · MongoDB Atlas Daily-use
🏙️ city-companion private Privacy-first mobility telemetry + AI survey generation for traffic analytics TypeScript · Expo · Express · Prisma Phase 4
🛣️ nextgen-highspeed-toll private Zero-hardware, fully simulated highway toll digital twin React · MongoDB · RabbitMQ · ROS 2 MVP
🚦 traffic-surveillance-system private Adaptive spatio-temporal traffic surveillance — helmet compliance, multi-frame wheel-count classification, real-time risk indexing Python · Computer vision Active
🧠 Harsh OS private Local-first 4-agent cognitive layer for LinkedIn — Tampermonkey ↔ Ollama, local vector DB, zero data leaves the machine Vanilla JS · Ollama v1
💼 DoppelDash private Internal operational platform built during the Doppelmayr internship TypeScript · React · Node.js · MongoDB Phase 14

🌍 Open-source contributions

Project Contribution
validatorjs/validator.js #2683 — fixed the uz-UZ mobile-phone regex, adding 5 missing carrier prefixes (33, 50, 55, 77, 88)
Textualize/rich #4123 — removed a dead link from the README
ToolJet/ToolJet Contributions to the open-source low-code platform

🏅 Awards & recognition

  • Smart India Hackathon 2025 — National Winner · first team from SIT Lonavala ever to win nationally · certificate
  • ISRO Robotics Challenge (IRoC-U 2026) — Finalist, Automation & Navigation Lead, Team ID 10144 · official listing
  • RoboRashtra 2026 — 1st Runner-Up, Chakravyuh maze & line solving (₹10,000)
  • AICTE & MIC — selected for Phase II of the IDE Bootcamp, Ministry of Education · certificate
  • Bentley Systems Illuminate 2026 — digital twin / MicroStation / OpenFlows summit participant
  • PressMaval Maza, state-level coverage
  • PCU Ideathon 2.0 — Top 40 of 500+ teams

Certifications — Certified Ethical Hacker Bootcamp (ZTM) · Advanced DSA (Codedamn) · Computer Networks (Udemy) · Web Developer Bootcamp 2025 (Udemy) · Data Science 101 (IBM)


🎓 Education

B.Tech, Computer Science — Sinhgad Institute of Technology (SIT), Lonavala · Sep 2024 – Aug 2027 CGPA 7.18 · Sem 5 SGPA 8.05 · O grade in Laboratory Practice I, CN & Security Lab, and Seminar & Technical Communication · Blockchain Honours: O (practical), A (theory) · A+ in Distributed Systems Coursework: Data Structures & Algorithms, Operating Systems, DBMS, Theory of Computation, Distributed Systems, Computer Networks

Diploma in Computer Engineering (First Class with Distinction) — SPM Polytechnic, Solapur · 2021–2024 · General Secretary ICSE Class 10 — Presentation Convent School, Nerul · 89 %

Targeting GATE 2027, top-500 rank — Distributed Systems, DBMS, Theory of Computation, Computer Networks.


✍️ Writing & speaking

  • dev.to/harshpandhe — 18 posts on robotics, system design, GitHub workflows, and Blender for autonomous systems
  • LinkedIn — engineering write-ups: drone-swarm safety architecture, the float('nan') truthiness bug that corrupted a trading NAV, learning-curve analysis before scaling a dataset, and the SSRN paper announcement
  • CODE2LAUNCH — workshop leader, 70+ attendees (React, Node.js, JavaScript)
  • Infotsav '25 — speaker (CityConnect)

📊 GitHub activity

Harsh Pandhe's GitHub contribution streak
Harsh Pandhe's GitHub trophies
Contribution graph rendered as a snake game

🚀 What's next

  • IRoC-U 2026 Finals — clear the final round with ISRO
  • Get the negative-result paper onto arXiv q-fin.PM (seeking endorsement)
  • RopeIQ — industry outreach and enterprise validation
  • ChronoTrack — enterprise pilot deployment
  • pod-transit Era 2 — finish the fleet / ROW / geo-export track
  • terra-verify — publish the validation methodology
  • GATE 2027 — top-500 rank
  • Speak at a robotics or edge-AI conference
  • ✅ Publish quantitative research to SSRN
  • ✅ Ship the Taramandal studio → GCS pipeline
  • MARS Phase 2 — CBF safety filter, collision avoidance, 100 % coverage
  • ChronoTrack v3.2.5 — 2FA, team-lead dashboards, auto-update installer
  • ✅ Open-source the maze-solver neural network firmware (Bot)

🤝 Contact

Open to: research collaborations · open-source work · robotics, autonomy and quantitative engineering roles. Particularly interested in: autonomous navigation · multi-agent systems · geospatial infrastructure · quantitative research · distributed systems.

Email Harsh Pandhe Book a call with Harsh Pandhe LinkedIn X (Twitter)

📍 Navi Mumbai, Maharashtra, India · 🌐 harshpandhe.com · 📧 harshpandhehome@gmail.com

Keywords: autonomous navigation engineer · multi-agent reinforcement learning · ROS 2 developer · PX4 drone software · LiDAR SLAM · Control Barrier Functions · quantitative research India · Nifty 500 momentum · Sentinel-2 remote sensing · ropeway alignment optimization · Harsh Pandhe

Pinned Loading

  1. codex-editor codex-editor Public

    A powerful, modern collaborative code editor built with Next.js, Monaco Editor, and Liveblocks. Designed for seamless real-time collaboration, perfect for pair programming, code reviews, team devel…

    TypeScript 1

  2. Project-Morpheus Project-Morpheus Public

    Official website for Project Morpheus 2026, a 24-hour hackathon hosted at STES Campus (SIT Lonavala & SKN). The site includes event details, problem statements, rules, and committees, with a cyberp…

    TypeScript 2

  3. TelhanSathi-SIH2025 TelhanSathi-SIH2025 Public

    Forked from dhirajdurgade7758/TelhanSathi-SIH2025

    Smart India hackthon 2025 project

    Jupyter Notebook

  4. harsh-pandhe harsh-pandhe Public

    🛠️ Computer Science @ SIT Lonavala | Architecting custom blockchains, training neural networks for autonomous robots, and building scalable full-stack solutions. Lead Organizer of Project Morpheus …

    1