Akshey Prabhakar
Machine Learning,Data Science
New Delhi
Summary
Accomplished Lead Machine Learning Engineer with a proven track record at Infinite Computer Solutions, enhancing user experience and retention through innovative AI solutions. Expert in model development and data analytics, with a knack for translating complex requirements into actionable insights. Demonstrates exceptional problem-solving and leadership skills, achieving significant performance improvements in high-stakes environments.
Overview
5
5
years of professional experience
4
4
Certifications
Work History
Lead Machine Learning Engineer
Infinite Computer Solutions
07.2021 - Current
- Engineered QoE analytics platform using decision tree regression and custom DTW-based forecasting. Processed 100K TPM, reduced buffering 45%, improved user retention 30%. Innovated with self-healing mechanisms, reducing manual interventions by 80%.
- Developed AI-driven buffer management system combining Gaussian mixture models for network simulation and LSTM for prediction. Achieved 50ms latency, 95% accuracy in underrun forecasting. Implemented novel unsupervised root cause analysis using DBSCAN clustering, cutting issue resolution time 70%.
- Architected real-time data processing pipeline with ASGI and Flink, handling 1M+ events/minute. Created custom windowing algorithm for efficient anomaly detection, reducing false positives 80%. Enabled near-real-time analysis of 50M+ daily records, cutting processing time from hours to minutes.
- Pioneered short-form video generation using fine-tuned LLMs and RAG on 10M+ content items. Implemented 650 FPS inference pipeline with custom attention mechanism for shot boundary detection. Increased user engagement 25%, reduced content creation time 80%.
- Designed scalable recommendation system using content-based filtering with custom vector embeddings. Implemented hybrid model with WARP loss and time-decay algorithms. Improved content discovery 35%, increased watch time 20%, achieving 40% better cold-start performance.
- Built multi-source sentiment analysis pipeline using zero-shot classification and fine-tuned BERT variants. Implemented active learning for efficient labeling. Achieved 90% accuracy across diverse platforms, reduced manual analysis time 60%.
- Developed advanced computer vision system for broadcast analysis. Integrated custom-trained models for logo detection, multi-lingual OCR (7+ Indian languages), and shot detection using modified TransNet V2. Improved automated metadata generation accuracy by 40%.
Lead Computer Vision(ML) Engineer
Darkhorse Sonador Inc
10.2020 - 06.2021
- Pioneered real-time soccer analytics pipeline, integrating custom-optimized YOLOv4 and DeepSORT. Tackled occlusion challenges with novel temporal consistency algorithms, achieving 95% mAP at 35 FPS on 4K streams.
- Engineered distributed inference system on AWS, overcoming latency issues with adaptive load balancing. Scaled to 10+ concurrent matches while maintaining sub-100ms end-to-end latency.
- Solved multi-camera calibration problem using RANSAC-based homography with non-linear refinement. Achieved
- Developed LSTM-CNN hybrid for action recognition, addressing class imbalance with focal loss and hard negative mining. Reached 98% accuracy on diverse, challenging soccer events.
- Tackled player re-identification in dynamic environments using Siamese networks with online hard example mining. Maintained 99% accuracy across varying lighting and occlusion conditions.
Co-Founder/ML Lead
BlackNet
07.2019 - 10.2020
- Engineered real-time smoke and fire detection system using IP camera feeds, achieving sensor-comparable accuracy. Developed custom CNN architecture, reducing implementation costs by 200% and expanding coverage to 100%.
- Innovated mask authentication and face recognition system using SSD and MobileNet with Radial Field Block. Optimized for edge deployment, achieving 15-20 FPS on NVIDIA Jetson Nano with 5x weight reduction using TensorRT.
- Architected scalable model serving infrastructure on AWS, supporting 10+ clients concurrently. Implemented face liveness detection for blockchain-based KYC, reducing fraudulent authentications by 16% using ensemble models with noise modeling and texture analysis.
- Leveraged DataBricks for ETL, extracting 15+ alternative KPIs for micro-financing models. Developed social distancing monitoring system with bird's-eye view transformation, achieving 95% accuracy in crowded industrial environments.
- Conducted model interpretability studies using occlusion experiments and GradCAM, leading to targeted improvements in face authentication models. Achieved 82% accuracy on CASIA Face Dataset for print and replay attack detection, outperforming baseline by 10%.
Education
Bachelor of Technology - Computer Science Engineering
Guru Gobind Singh Indraprastha University
Delhi 04.2001 -
12th - Non-Medicine Science
Cambridge Foundation School
West Delhi 04.2001 -
Skills
Natural Language Processing
Model Development
Data Analytics
Machine Learning
Supervised Learning
Machine Learning Integration
Feature Engineering
Clustering Algorithms
Gradient Boosting Machines
Statistical modeling
Neural Networks
Programming Language: Python, CTools and Frameworks: Pytorch, VOTT Dataset Annotation, AWS, Flask, Tensorboard, MLFLOW, Albumentation, Netron, Scikit Learn, Dash ,Pandas, Tflite, Tensorflow, Keras, DataGrip, Dash, HuggingFace , Pytorch Lightning,Data Science and Analytics : Pandas, Vaex, Matplotlib, Seaborn, Plotly, NLTK, SpacyBI Tools: Google Analytics , PowerBI , Tableau, Azure AnalyticsCloud Skills : AWS, GCPAI/ML Data Modelling, Data Mining, Big Query(Data Warehouse)
Certification
WorldQuant University Applied Data Science Certification
Accomplishments
- Among Top 10 in Rank Analytics Vidhya Computer Vision Competition
- Bertelsmann Scholarship 2019
- WorldQuant University Scholar
- 1st Runner-Up UltraHack
- Employee of the Quarter
- Spark of Infinite
Timeline
Machine Learning and Statistical Analysis Certification(WorldQuant University)
08-2024
Lead Machine Learning Engineer
Infinite Computer Solutions
07.2021 - Current
Lead Computer Vision(ML) Engineer
Darkhorse Sonador Inc
10.2020 - 06.2021
WorldQuant University Applied Data Science Certification
06-2020
Google Explore ML Facilitation
12-2019
Co-Founder/ML Lead
BlackNet
07.2019 - 10.2020
INTEL AI From the Data Center to the Edge,An Optimized Path Using Intel Architecture and OpenVINO
06-2019
Bachelor of Technology - Computer Science Engineering
Guru Gobind Singh Indraprastha University
04.2001 -
12th - Non-Medicine Science
Cambridge Foundation School
04.2001 -