Amit Dedhia
Deep Learning Engineer - Computer Vision/NLP
Mumbai, Maharashtra 400092--
Looking for remote position (based in India) from companies across the world. Have over 2 decades of
work experience in Software development and architecture. Now working in the most exciting field of
AI (Computer vision and NLP).
Work Experience
Independent Contractor - Machine Learning
Canopy - Remote
January 2022 to Present
Description:
Improve accuracy and performance of existing deep learning models for bank code detection and sub
type detection for a PDF statement received from the clients.
Dev Tools: PyTorch, OpenCV
Team size: 1
Domain: Fintech
Role:
• Go through existing implementation of Bank code detection and test performance
• Implementation of CNN based neural network to identify bank type from financial statement
• Data preparation
• Model development, training and evaluation
Machine Learning Developer
viAct.ai - Remote
May 2021 to October 2021
Description:
Development of product to improve workplace safety and increase productivity in construction industry,
by applying proprietary AI vision technologies.
Dev Tools: PyTorch, OpenCV
Team size: 5
Domain: Construction
Role:
• Code Implementation and training for various object detection and classification models
• SSD - https://github.com/Amitdedhia6/SSD_PyTorch
• SSD - https://github.com/Amitdedhia6/DeepLearningProjects1/tree/main/core/models/detection/ssd
• SSD Lite https://github.com/Amitdedhia6/DeepLearningProjects1/tree/main/core/models/detection/ssd_lite
• Efficient Net https://github.com/Amitdedhia6/DeepLearningProjects1/tree/main/core/models/classification/
efficientnet
• Darknet53 https://github.com/Amitdedhia6/DeepLearningProjects1/tree/main/core/models/classification/darknet
• YoloV3 https://github.com/Amitdedhia6/DeepLearningProjects1/tree/main/core/models/detection/yolov3
https://github.com/Amitdedhia6/DeepLearningProjects1/blob/main/core/criterions/yolov3_loss.py
• Multi GPU training https://github.com/Amitdedhia6/DeepLearningProjects1/blob/main/train_classification.py
• Research on digital measurement - 3D reconstruction, Structure from motion, Tools (Visual SFM,
MeshLab, CMVS)
• Prototype - identifying traffic jam conditions on road
• Prototype - person tracking from multiple cameras at a site
Machine Learning Developer
Aerobi - Remote
June 2020 to May 2021
Description:
Aerobi AI-powered software is a fitness app allows user to see herself on TV while working out at home,
get real-time feedback on form and performance to get truly immersive experience.
Dev Tools: TensorFlow, PyTorch, OpenCV
Team size: 5
Domain: Fitness
Role:
• MVP Product development.
• Computer Vision, Object detection and Semantic segmentation Neural Network models such
as Mask RCNN, ResNet, MobileNetV2, Shufflenet, UNet, Single Shot Detector
• Quantization, Model Compression (AIMET libraries, TensorFlow)
• Pose Estimation - OpenPose, Custom model development (using UNet and BlazePose model concepts)
- https://github.com/Amitdedhia6/BlazePose
• Google Chrome Extension Development using BodyPix TF-JS model, Google Analytics, Auth0
https://chrome.google.com/webstore/detail/aerobi-enhance-youryoutu/mlfkmhibffpoleieiomjkekmjipdekhg?hl=en
• VIA annotation software customization.
• YouTube video recommender system - a prototype recommender system aimed to provide better
search results for fitness related videos https://github.com/Amitdedhia6/YouTubeFitnessRecommenderSystem
Deep Learning Intern
Pucho Technology Information Private Limited - Remote
March 2020 to June 2020
• Identification of new molecules for Drug Discovery
Identification of new molecules for drug discovery. Implemented using Pytorch, using GAN having
RNN for discriminator and also for Generator. This project is more of research and experimentation.
https://github.com/Amitdedhia6/DrugDiscovery
• Flight Delay prediction
Prediction of flight delay using data from Bureau of Transportation Statistics. Used Reporting Carrier
On-Time Performance data for year 2016 to generate the model.
https://github.com/Amitdedhia6/FlightDelay
• Amazon bidding optimization
A prototype application (as proof of concept) based on very limited availability of data to predict
expected ACoS (Advertising Cost of Sale) for a given input keyword.
https://drive.google.com/open?id=1ryUnoPOKDQ0LbzcQ5kOlmmP7KApQUV_H
PRIOR EXPERIENCE (Not Involving Deep Learning and Machine Learning)
Consultant (Part time)
Liquidity Offset Network - Remote
October 2019 to January 2020
Product: OTSafe - a central counterparty service for OTC traders
Domain: FinTech
Role: Understanding OTSafe product requirements, database design and technical documentation for
product development facilitation (writing detailed technical workflow), testing.
Senior Engineer
Canopy - Singapore
September 2015 to December 2019
Product: Canopy - Wealth Management software for better insight and holistic view
Domain: FinTech
Role: Architected and developed (end to end) complex modules for Canopy product's Engine subsystem:
modules such as CanopyUL (Universal Language), EOD Processing, Stashboard
(Statement Processing Dashboard), Blotter system, Real time data migration for analytics system, Access
control, Data feeds processing from various customer data such as Bank of Singapore, Julies
Baer, Corecam etc.
Dev. Tools: Python 3.7, VB.NET, Postgres, JavaScript
Senior Solution Architect
CitiusTech - Mumbai, Maharashtra
January 2013 to September 2015
Mumbai. (Jan 2013 ~ Sep 2015)
• Client: FujiFilm Medical Systems, Leading PACS vendor in USA
Product: Multi Modality Medical imaging workstation software
Domain: Medical Imaging, DICOM
Role: Delivery lead for a team 25+, Design and Architecture, Requirement gathering and documentation,
Code reviews, Verification and Validation, Overall project management ensuring timely delivery and
process adherence, Customer Management, People Management.
Dev. Tools: .NET 3.5, C++.
Technical Architect
Blue Star Infotech Ltd - Mumbai, Maharashtra
March 2003 to January 2013
• Client: CoreLab Partners, USA
Product: Medical imaging workstation software for clinical trials
Domain: Medical Imaging, DICOM
Role: Presales, Requirement understanding (domain expert), Validation and Verification of Image review
application, Project Management, Customer Management, People Management.
Duration: 22 Months (Mar 2011 – Jan 2013)
Achievements: Received the best customer satisfaction rating (10/10); highest in the company for year
2011. Survey was conducted by a third-party market research company.
• Client: Sunquest Info, USA
Domain: Laboratory Information Systems
Role: Presales, Software Architect, Tech Lead, Developer
Duration: 12 Months (Feb 2010 – Feb 2011)
Dev. Tools: .NET 3.5, WPF, Visual C++ 6.0, VB.NET, Visual Basic 6, Codejock Skinning tool
• Client: Stratagene, USA (An Agilent Technologies company)
Product: Prius, workstation software for QPCR (Quantitative Polymerize Chain Reaction) machine
Domain: Molecular Biology, Genetics
Role: Software Architect, Project Manager, Developer r
Duration: 21 Months (Oct 2007 – Feb 2010)
Dev. Tools: .NET 3.5, WPF/XAML, FxCop, NUnit, Enterprise Architect
• Client: MedicSight PLC, London, UK
Product: MedicRead - a Computer Aided Detection (CAD) product for CT colonography
Domain: Medical Imaging, DICOM
Role: Medical Imaging, DICOM
Duration: 2 Months (Aug 2007 – Sep 2007)
• Client: Stratagene, USA (An Agilent Technologies company)
Product: MxPro, workstation software for QPCR (Quantitative Polymerize Chain Reaction) machine
Domain: Molecular Biology, Genetics
Role: Tech Lead, Developer, Onsite coordination
Duration: 7 months (May 2006 – Nov 2006)
Dev. Tools: Visual C++ 2005, MFC, C++ CLI, XML, XSL.
• Client: Hitachi Medical Corporation, Japan
Product: Multi Modality Console, a multi modality medical imaging workstation software
Domain: Medical Imaging, DICOM
Role: Software Architect, Tech Lead, Developer, Onsite coordination.
Duration: 3 years and 1 month (Mar 2003 – Apr 2006)
Dev. Tools: .NET XDE, C#, Managed C++, Cedara IAP (Image Analysis Platform) Toolkit, ADO.NET, DICOM,
Rational Quantify, .NET Memory Profiler, ANTS and DevPartner profilers.
Project Leader
Corcomp Infosystems Ltd - Mumbai, Maharashtra
March 2001 to March 2003
• Development of an End to End N Tier Business application for Apparel retailing that covers Product
Setup, buying and merchandising, Sales and Distribution, Pricing and Schemes management, Point of
Sale (POS), Planning and MIS.
Technologies: .NET 1.0, C#, ADO.NET, SQL Server 7.0, COM+
• An N Tier business application automating the general operations, personnel operations and financial
accounting of the client. The client is in the business of road transportation of goods to and from the
Bombay port and within the country.
Technologies: .NET 1.0, C#, ADO.NET, SQL Server 7.0, COM+
• Development of a multi-threaded framework to develop applications using Intelligent, talkative and
animated character for enhanced human interface.
Technologies: C++, Win 32 SDK Programming, COM, MS Agent character APIs, XML
Senior Systems Engineer
Tata Infotech Ltd
July 2000 to November 2000
• Development of Optical test system software for Mania Technologie, Germany. Worked as Senior
Developer in a team of 2. Project duration was 4 months (from Jul 2000 to Nov 2000). Designed algorithms
to determine offset in an image of a Printed Circuit board and to determine the error regions.
Technologies: C++, Windows SDK programming, ADO and SQL Server 7.0
• Development of a client server application called OSMOS (Offline Surveillance and MOnitoring System)
for Reserved Bank of India, Mumbai, India. Worked as Software Programmer in a team of 20. My
involvement in the project was for 4 months (from Mar 2000 to Jun 2000).
Technologies: Visual Basic 6.0 with MS Access 97 and Oracle 8.0 as backend database.
• Development of a document management product called FileIT. The product features include scanning,
indexing, storage and retrieval of documents; online and offline archival of documents; batch scanning
of documents; access rights to documents. Worked as Team Leader for a team of 9. Project duration was
1 year and 4 months (from Nov 1998 till Mar 2000).
Technologies: Visual C++ 6.0, MFC, ATL, Visual Basic 6.0, MS Access 97, MS SQL Server 6.0
Trainee Programmer (part time)
Siemens Information Systems Ltd - Mumbai, Maharashtra
October 1997 to July 1998
• (Academic Project) Development of Attendance Monitoring and Leave Maintenance System during final
academic year in engineering. Worked as a Trainee Programmer (part time) in a team of 4. Development
in Visual C++ 5.0, MFC, DAO and MS Access 97 as the database.
Education
Natural Language Processing with Deep Learning
Stanford Center for Professional Development
February 2022 to April 2022
Mathematics for Machine Learning
Imperial College London - Coursera
November 2021 to January 2022
Bachelor's in Computer Science
University of Mumbai - Mumbai, Maharashtra
August 1994 to July 1998
Skills
• Computer Vision and NLP using deep neural networks (2 years)
• Model development, Training and Evaluation (2 years)
• PyTorch (2 years)
• TensorFlow (1 year)
• Python (5 years)
• OpenCV (2 years)
• .Net (10+ years)
• OOAD, UML, Architectural and Design patterns
• Design of Multi-threaded applications
• Postgres (3 years)
• SQL
• JavaScript
• Git
• XML
• APIs
• REST
• Computer Science
• GitHub
• Software Development
• Application Development
Languages
• English - Fluent
Links
https://www.linkedin.com/in/amit-dedhia-
Certifications and Licenses
Natural Language Processing with Deep Learning
April 2022 to Present
https://digitalcredential.stanford.edu/check/
B4292DEC39FA2D68F6FB28D65C88C28FB36B2464A-DCB0BBF3E433814FSUdqSzZVbVVjQmZSVktydFhZTVIzcU
Received from Stanford Center for Professional Development
Mathematics for Machine Learning
January 2022 to Present
Online non-credit course offered by Imperial College London and Coursera
https://www.coursera.org/account/accomplishments/specialization/certificate/ZU2UBHPS226C
