A
Aravind Kandiah
About
Detail
CTO & Co-Founder
San Francisco, California, United States
Résumé
Jobs
verified_user
0% verified
CTO & Co-FounderBifrost AIFeb 2020 - Current (6 years 4 months)Backed by Sequoia, Lux Capital, Airbus Ventures, Carbide, Wavemaker. Bifrost recreates the world in simulation. Our infrastructure powers the world’s largest robotics companies to train and evaluate physical AI at the speed of software, from autonomous Mars exploration with NASA to automating high-risk industrial work and reducing loss of human life.- AI ResearcherMedios TechnologiesJan 2019 - Dec 2020 (2 years)
Education
verified_user
0% verified
- Singapore University of Technology and Design (SUTD)Singapore University of Technology and Design SUTD
Projects (professional or personal)
verified_user
0% verified
- IIdentity Verification Mobile ApplicationFeb 2018 - Apr 2018 (3 months)Developed a mobile application that is able to verify identity through facial feature matching between passport photos and selfie. Application was built in partnership with IBM. Built using Ionic, angularJS, OpenCV, Tensorflow
- TTraffic Sign ClassificationJan 2017 - Dec 2017 (1 year)Built and trained a deep neural network to classify traffic signs, using TensorFlow. Experimented with different network architectures and developed data augmentation pipeline to increase accuracy and robustness against varying environmental conditions without obtaining additional data. Performed image pre-processing and validation to guard against overfitting.
- VVehicle Detection & TrackingJan 2017 - Dec 2017 (1 year)Created a vehicle detection and tracking pipeline with OpenCV, histogram of oriented gradients (HOG), and support vector machines (SVM). Optimized and evaluated the model on video data from an automotive camera taken during highway driving.
- AAdvanced Lane FindingJan 2017 - Dec 2017 (1 year)Built an advanced lane-finding algorithm using traditional computer vision algorithms such as distortion correction, image rectification, color transforms, and gradient thresholding. Resulting in an application that is able to Identify lane curvature and vehicle displacement from a single dash-camera. Algorithm is able to detect changes in the environment and dynamically alter itself to be robust against changing road and weather conditions. Temporal smoothing was developed for increased stability.
- BBehavioural CloningJan 2017 - Dec 2017 (1 year)Developed a CNN to map raw pixels from a single front-facing camera directly to steering commands. Model was trained and tested using a simulator built in Unity. This end-to-end approach was inspired by Nvidia’s paper on End to End learning.
- PProject TriggeredJan 2016 - Dec 2016 (1 year)Lead software developer of a team of 5. Utilizing computer vision algorithms to track the movement of people and objects through space on computationally limited systems. With this data we were able to draw virtual tripwires to trigger customizable actions as we move through space and place virtual buttons in space. Awarded BP Book Prize.
- DDigitising Physical Human ExpressionJan 2015 - Dec 2015 (1 year)Led a team of 5 in the creation of the Zen Smart pen. A smart pen that digitises whatever you write or draw on any surface & autocorrects spelling. In order to benefit from the many splendours of the virtual medium while still preserving the tactile and natural physical expression of writing. *Received A Grade
- PPersonalised Virtual Assistant & Data AggregationJan 2015 - Dec 2015 (1 year)Led a team of 4 in building a smart mirror. The smart mirror is able to recognise individuals through facial recognition & display personalised content such as deadlines & upcoming events that matches individuals interest. *Received A Grade
Awards
verified_user
0% verified
- 1st Place, Amazon Machine Learning CompetitionAmazonJan 2018Emerged as grand champions of the competitions and was selected to represent Singapore to race self-driving cars in Jakarta. We built a transcription and topic modeling application. It can transcribe any meeting from speech audio, summarise what was discussed and split up the conversation based on the person and topics. This enabled us to create meeting note summaries on the fly while a meeting was happening and send them to everyone at the meeting.
- 2nd Place, Amazon Self-driving Car ChallengeAmazonJan 2018We represented Singapore to race self-driving cars in Jakarta. By Utilising an End to End trained neural network along with a PID controller we were able to place 2nd place. We utilized Tensorflow to build the models and trained them on AWS Sagemaker.
- C2nd Place, Learning Innovation ChallengeCivil Service CollegeJan 2018Won 2nd place & people's choice award. Built an application that crawled through their entire video database, transcribed all the video, modeled the content of each lecture by topic. So when a user searches for a certain topic we can create courses on the fly across different lectures so that the user can learn exactly what they want to learn in the quickest possible time.
- P1st Place, Pitbull's Smackathon Blockchain CompetitionPitbullJan 2018Emerged as Grand Champions in this month-long international competition. We built a decentralized stock exchange for musicians on the Ethereum network. Fans can buy into their favorite artist and bet on their success and artist can get additional funding through this mechanism. Built using React and Ethereum
- Hack4Climate ChampionsUnited NationsNov 2017100 technologist from around the world were selected and flown into Germany to compete. We emerged victorious and presented our blockchain & mobility optimisation product at the COP23 United Nations Conference in Bonn,Germany
- HPeople's Choice AwardHypermotion GermanyNov 2017Voted best project by the community. Developed a package measurement and optimisation solution for the logistics industry using augmented reality and novel bin-packing algorithms
- H1st Place, Global VR HackathonHTC ViveJul 2017Won 1st place by developing a Virtual Reality application for HTC Vive and Samsung Gear. The application brings massively scalable user experience design testing to the architecture and construction industry. Allowing architects and engineers to make data driven design decisions to aid them in building better physical and virtual spaces in a shorter time. Link to our presentation we gave at SGinnovate: https://www.youtube.com/watch?v=8nSTVzvZbBM&t
- 2nd Place, Microsoft AzurenautsMicrosoftJan 2017Using recurrent neural networks to classify human behavioral patterns in real time.
- 1st Place, Hood Disrupt 2017JLLJan 2017Grand prize winner for building a autonomous spatial mapping robot for construction sites. Solution is able to improve real time communication on/offsite , document site progress and help speed up conflict resolution. Resulting in higher productivity on the work site. Technology Stack: Computer Vision and Spatial Mapping Virtual Reality Robotics
- British Petroleum Book PrizeBritish PetroleumSep 2016Awarded the book prize and prize money for Project Triggered.
- hAEC Singapore WinnershttpsaechackathoncomaboutusJan 2016Won the hackathon, using augmented reality and spatial mapping on Hololens to resolve on-site and CAD model conflicts.
- SColours AwardSingapore Schools Sports CouncilFor excellence in academics & sports
- 2nd Place and best in category prize for the SIA App ChallengeSingapore AirlinesDeveloped an Augmented Reality solution that enabled Singapore Airline Cargo operators to quickly scan packages and obtain precise dimensions in a matter of seconds without the need for additional hardware. With the dimensions, the solution was also able to optimise the way the packages should be packed into the cargo hold to maximise space and increase profits. Technology Stack used, OpenCV ,Python ,AWS Lamda, iOS Swift
- RRaffles DiplomaThe Raffles Diploma (RD) is a certification that recognises student participation and achievements in various domains such as Cognitive, Character and Leadership, Community and Citizenship, Arts and Aesthetics, and Sports and Health.