Real-Time Computer Vision Engineer

Abhishek Kumar Email: ******************@*****.***

Github: github.com/abhi47146 Mobile: +91-778******* Education

•

Indian Institute of Technology (BHU), Varanasi Varanasi, India Bachelor of Technology - Electronics Engineering 2025 Skills Summary

• Languages: C/C++, Python, SQL

• Domains: Software Development, OOPs, DSA, System Design, Machine Learning, Computer Vision

• Platforms: Linux, NVIDIA Jetson, Raspberry Pi, GPUs

• Frameworks and Tools: FastAPI, Flask, PostgreSQL, WebSockets, OpenCV, YOLO, ByteTrack, TensorFlow, Pytorch, Docker, LLM, CVAT

Experience

• Associate - ITS Planners and Engineers: June 2025 - Present

Developed a real-time vehicle detection, classification, and counting system using YOLO and ByteTrack to monitor traffic flow from CCTV footage across multiple vehicle classes.

Built a Live Tracking View and Live Model Testing View using FastAPI and WebSockets, enabling users to connect any RTSP stream and observe real-time model inference directly in the browser.

Developed a real-time Vehicle Monitoring Dashboard displaying detected vehicle number plates, vehicle speeds, and corresponding vehicle snapshots for efficient traffic monitoring and event tracking.

Designed and implemented queue length estimation logic using pixel-coordinate analysis to detect vehicle queues and measure congestion levels accurately.

Integrated MediaMTX for live camera streaming and for continuous NVR-style video recording and surveillance management across multiple CCTV streams.

Built an adaptive traffic light automation system that dynamically adjusts signal timings based on real-time vehicle queue detection and traffic flow data.

Deployed ML models for vehicle classification and counting on edge devices including NVIDIA Jetson and Raspberry Pi, optimizing models for low-latency inference on constrained hardware.

Developed a multi-parameter accident detection algorithm using OpenCV, combining per-vehicle speed estimation, bounding-box collision detection, stationary vehicle counting per frame, and vehicle angle-change tracking to identify abnormal driving behavior.

Annotated large-scale video datasets with bounding boxes across all vehicle classes to support computer vision model training on CVAT.

Incorporated fire and smoke detection into the same pipeline to flag hazards from live video feeds, and designed a weighted scoring system (tuned threshold values across all parameters) to trigger real-time accident alerts while minimizing false positives.

Tech Stack: Python, YOLO, ByteTrack, OpenCV, Computer Vision, FastAPI, WebSockets, MediaMTX, Kerberos.io.

• Technical Intern - Careers360: May 2024 - July 2024

Solved competitive programming, mathematics, and Data Structures and Algorithms (DSA) problems.

Authored technical articles explaining key DSA concepts and problem-solving approaches.

Debugged solution errors and provided clear, structured explanations for improved understanding.

Tech Stack: C/C++, Content Management Systems (CMS), Microsoft Excel, MS Word. Projects

• Diagnosis of patient’s lung condition - detection of crackles and wheezes — Mosaic’23: Developed a machine learning model to classify lung conditions into wheezes, crackles, both, or none using lung sound recordings. Converted audio signals into spectrogram images using the Librosa library for feature extraction and visualization. Preprocessed audio data and mapped each recording to its corresponding annotation file. Designed and trained a Convolutional Neural Network (CNN) using Keras with TensorFlow backend for automated feature learning and classification. Exposure: Python, Keras, TensorFlow, CNN

• Deep learning model for image inpainting — Mosaic’23: Trained a neural network to predict missing parts of an image such that the predictions are both visually and semantically consistent. Used a CIFAR10 dataset to train the autoencoder decoder model and test it on the given dataset. Exposure: Python,Keras and TensorFlow, CNN Honors and Achievements

• Specialist on codeforces with max rating of 1436. ID: codeforces.com/abhishek47146

• 3-star on codechef with max rating of 1741. ID: codechef.com/abhishek47146

• Secured Global Rank 304 in Google Kick Start Farewell Round C 2023.

• Secured Global Rank 1 in July Long Two 2022 Division 4 on codechef.

• Winner of EPIIC Book Review Contest Jan 2026 held all over India.

• Cleared Jee Advanced 2021 with AIR 6289.