Ashish Umrao

Ashish Umrao

LocationPune, India
Phone+91 9950247507
3
Months Experience
2
Featured Projects
22
Technical Skills

Work Experience

Software Developer Intern

Haqdarshak (C4GT DMP)
2025-06 - 2025-08
  • Developed a Beckn-compliant BPP in Node.js, making a 15+ course catalog discoverable on open networks and achieving 100% compliance on core search/select actions.
  • Boosted search accuracy by over 75% by integrating Elasticsearch, enabling distinct, schema-aware queries for both courses and certifications.
  • Cut debugging time for asynchronous flows by 90% by architecting and deploying a complete 4-service Beckn sandbox (BAP, BPP, Gateway, Registry) with Docker Compose.

Education

Bachelor of Engineering - Information Technology

Army Institute of Technology
Timeline2024 - 2028
GPA / Grade8.8

Skills & Technologies

CC++JavaScriptPythonJavaBash

Featured Projects

IoT Sentry (eRaksha)

IoT Sentry (eRaksha)

Architected an autonomous security agent using LangGraph to implement a continuous Observe-Reason-Act loop that discovers IoT devices, audits vulnerabilities via the NVD API, and auto-isolates threats - without human intervention.

Engineered a custom Random Forest classifier trained on 13,904 real network flows; achieved 98% overall accuracy (F1: 0.98–0.99) in distinguishing normal traffic from DDoS, SYN flood, and brute-force attacks using bidirectional packet and byte features.

Built a multi-layer threat pipeline: Scapy for real-time ARP/DHCP sniffing, Nmap for port/OS fingerprinting, Cowrie honeypots for attacker deception, and ARP spoofing for automatic device quarantine.

Deployed the full stack via Docker Compose with a Streamlit dashboard exposing live risk scores, intrusion logs, and behavioral anomaly alerts to a unified web UI.

LangGraphPythonScapyNmapScikit-learnDocker
View Project
Fish Catch Identification

Fish Catch Identification

Secured 1st Rank at Smart India Hackathon 2025 for developing an offline-first Android application for real-time fish catch analysis for ISRO.

Engineered an on-device Edge AI pipeline using YOLOv8-Tiny + EfficientNet-Lite achieving 97% accuracy in species identification and freshness detection without internet connectivity.

Used ARCore depth sensing and reference scaling to implement real-time volume/weight estimation; developed a local-first SQLite architecture for geo-tagged record storage in low-connectivity coastal environments.

Deep LearningAndroidPythonTensorFlow
View Project

Achievements & Awards

1st Runner Up - eRaksha Hackathon IIT Delhi

IIT DelhiSecured first runner up at the national-level eRaksha Hackathon for AI Impact Summit 2026.
Awarded2026-01

Smart India Hackathon 2025 Winner

ISROSecured first place in the national-level Smart India Hackathon for the ISRO-sponsored challenge.
Awarded2025-12

Dedicated Mentoring Program Selection

C4GT DMPOne of 94 selected for Code for GovTech's DMP 2025 from over 4,800 submissions.
Awarded2025-06

Finalist - Adobe India Hackathon 2025

AdobeSecured a place in the Top 100 out of over 260,000 participants.
Awarded2025-10

Winner - Nirmaan 3.0 Hackathon 2025

MIT AOE, PuneWon 1st prize in a National Level Hackathon at MIT AOE campus.
Awarded2025-04

Grand Finalist - Innerve 9.0 Hackathon

AIT, PuneOne of the top 31 teams from 1,200 competing teams.
Awarded2025-02