Arpan Shingala

- Refined image-processing models that detect known and unknown logos and extract predefined invoice details using low-resource techniques, deployed as a local pipeline accessible through APIs and a user interface. - Developed automation systems for real-time applications by programming PLCs and provided support for an industrial client of Silicon Bits. - Developed a real-time laser-based distance measurement system using centroid detection and triangulation, improving processing speed from 3 to 30 FPS thro

- Designed and built a secure trading platform with authentication, encryption and real-time monitoring, including protections against common cyberattacks. - Worked directly with clients to refine features and meet their requirements while improving scalability. - Stack: FastAPI, Express.js, React/Next.js

- Improved a Punjabi acoustic model by 32% in accuracy by integrating new data in real time into the SRT subtitle workflow. - Built a UI tool that calculates transcription accuracy automatically and fixed line-break handling for consistent SRT output. - Evaluated and fine-tuned OpenAI Whisper (offline) for Punjabi speech-to-text.

- Built a learning-based fuzzing tool for gQUIC that infers the protocol's state machine (L* algorithm, modified quic-scapy for gQUIC 43) and uses it to guide testing. - Detected a known critical vulnerability, which validated the approach against real bugs. - Discovered previously undocumented behaviour in the lsquic library. - Stack: Python, quic-scapy, Wireshark.

- Carried out web application security assessments on real-world applications using tools such as Burp Suite. - Identified and reported vulnerabilities and verified fixes.

Assessed methods for analysing QUIC implementations, focusing on vulnerabilities and performance bottlenecks. Proposed and validated bug-detection techniques.

- Organised technical workshops and hackathons. - Promoted open-source tools and onboarded new members.

Focus: IT security and cryptography. Research interests: network protocol security, applied cryptography, learning-based testing.

Relevant coursework: Data Structures & Algorithms, OOP, DBMS, Linear Algebra, Mathematical Statistics, Optimisation, Machine Learning Undergraduate research: DRDO-sponsored work on QUIC protocol fuzzing (see Experience). Projects: learning-based fuzzing of QUIC implementations, AST-based static a