Chandrachud Basavaraj, Developer in Paris, France
Chandrachud is available for hire
Hire Chandrachud

Chandrachud Basavaraj

Verified Expert  in Engineering

Machine Learning Developer

Location
Paris, France
Toptal Member Since
May 15, 2018

Chandrachud is an experienced machine learning (ML) and artificial intelligence (AI) engineer. He studied at top schools in India and France and was, until recently, CTO of an early-stage AI startup. Chandrachud loves ML, computer vision, and large language models (LLMs) and has delivered high-quality work on various ML projects. He has been building and integrating multiple components to deploy AI systems in production on the cloud. Chandrachud loves to innovate and create cool technology.

Portfolio

StillMind (Freelance Projects)
Generative Pre-trained Transformers (GPT), Natural Language Processing (NLP)...
Mo-ka SAS
PyTorch, OpenCV, Socket.IO, Raspberry Pi, Flask, GStreamer, Machine Learning...
SkillCorner
Deep Learning, TensorBoard, TensorFlow, OpenCV, Python, Machine Learning...

Experience

Availability

Part-time

Preferred Environment

Bitbucket, PyTorch, Sublime Text, Ubuntu, Google Cloud, Python

The most amazing...

...deep neural networks I've worked on include Mask R-CNN for real-time object detection and segmentation and open-weight LLMs for uncensored language generation.

Work Experience

Indie Machine Learning Engineer

2017 - PRESENT
StillMind (Freelance Projects)
  • Trained an object detection and segmentation model on a screening mammography dataset to detect breast lesions and classify them as benign or malignant.
  • Developed a gun and human detection system for a client in the security industry to deploy in schools, malls, and other public places. The system allows early detection of potential shootouts and subsequent intervention by law enforcement agencies.
  • Created a Telegram chatbot with both text and voice capabilities, using an uncensored LLM (large language model) similar to ChatGPT but with fewer restrictions.
  • Integrated a facial key point detection system with some bespoke elements and a web front end for a VR/AR application in the fashion industry.
  • Developed a computer vision system for parsing building floorplans (images, PDFs, other formats) and recognizing walls and rooms for downstream use in wireless network planning.
  • Built a motion detection system based on background detection that works even against dynamic backgrounds (e.g., moving water, tree leaves, and more) based on state-of-the-art algorithms in the scientific literature.
  • Built a scene text recognition pipeline using two deep neural networks, EAST for detection and ASTER for recognition, with fast and reliable performance.
  • Contributed to aspect-based sentiment analysis using classical NLP methods and spaCy and a deep learning approach using BI-LSTM-CRF implemented in PyTorch.
  • Adapted and combined a fast implementation of an RNN using Simple Recurrent Unit (SRU) with relational networks for question answering on Facebook Research's bAbI tasks, with excellent results.
  • Collaborated on various other confidential projects for clients in computer vision, natural language processing, recommender systems, and other classic machine learning areas.
Technologies: Natural Language Processing (NLP), Generative Pre-trained Transformers (GPT), Google Cloud, MATLAB, SpaCy, Keras, TensorFlow, PyTorch, Scikit-learn, Python, Machine Learning, Artificial Intelligence (AI), Computer Vision, Image Recognition, OpenCV, NVIDIA CUDA, Chatbots, Generative Artificial Intelligence (GenAI), Convolutional Neural Networks (CNN), CAD, Google Maps SDK, AI Programming, Large Language Models (LLMs), Audio, Bots, Language Models, Linux, MacOS, Weka, Full-stack, Deployment, Google Cloud Platform (GCP), APIs, Neural Networks, Data Science, Software Architecture, Algorithms, JavaScript, Bash, DevOps, Prompt Engineering, Full-stack Development, NumPy, Back-end, Generative Pre-trained Transformer 3 (GPT-3), OpenAI GPT-4 API, BERT, Docker, GPU Computing, Technical Architecture, Back-end Architecture, LangChain, Pinecone, Applied Research, Training, Fine-tuning, Transformer Models, You Only Look Once (YOLO), Computer Vision Algorithms, LlamaIndex, Cloud Computing, Cloud, AI Chatbots, Data Modeling, Databases, Snowflake, Artificial Intelligence (AI), Image Recognition, Machine Learning, Project Management, Project Coordination, Linux Servers, Twilio, Python Asyncio, FastAPI, PDF, Object Detection, Object Tracking, Generative AI, Data Engineering, Video Processing

CTO | Computer Vision and AI Engineer

2018 - 2023
Mo-ka SAS
  • Built the 1st proof of concept to turn the founder's idea into a startup with angel investors. This included computer vision, system design, and implementation. Continued to head the technical team going into commercialization.
  • Developed the company's first product, a smart grocery basket that uses computer vision and ML to recognize the products added and removed by a customer and automatically compute the grocery bill.
  • Headed the development of the company's next product after a pivot, a supermarket self-checkout monitoring and fraud detection system using computer vision, machine learning, and a stateful event detection system.
  • Trained deep learning object detection and segmentation models with data augmentation and hyper-parameter tuning to obtain high accuracy. Conceived and built a highly modular and flexible computer vision pipeline for real-time video processing.
  • Developed systems for deploying deep learning models in production using a Python, Flask, and Vue web stack, OpenCV and PyTorch for computer vision, and other confidential elements.
Technologies: PyTorch, OpenCV, Socket.IO, Raspberry Pi, Flask, GStreamer, Machine Learning, Artificial Intelligence (AI), Computer Vision, Python, Image Recognition, TensorFlow, Image Processing, NVIDIA CUDA, Convolutional Neural Networks (CNN), Architecture, Integration, AI Programming, Linux, Amazon Web Services (AWS), FFmpeg, Vue, Full-stack, Deployment, Google Cloud Platform (GCP), APIs, Neural Networks, Data Science, Software Architecture, Algorithms, REST APIs, JavaScript, Leadership, Bash, DevOps, AWS DevOps, SaaS, Full-stack Development, Node.js, Express.js, TypeScript, NumPy, Team Leadership, Project Management, Machine Learning Operations (MLOps), Back-end, modal, Docker, GPU Computing, SQL, Redis, Celery, OpenAPI, NoSQL, Technical Architecture, Solution Architecture, Back-end Architecture, Applied Research, Training, Fine-tuning, Computer Vision Algorithms, Cloud Computing, Cloud, Data Modeling, Databases, Artificial Intelligence (AI), Image Recognition, Machine Learning, Project Management, Project Coordination, Linux Servers, Parallel Computing, Multithreading, WebRTC, Object Detection, Object Tracking, Data Engineering, CTO, Video Processing

Deep Learning Intern

2017 - 2017
SkillCorner
  • Built a high-accuracy real-time object detection system for detecting players and the ball in live football video using state-of-the-art deep neural networks.
  • Developed a high-accuracy multi-digit number recognizer for player jersey numbers using transfer learning from the SVHN dataset.
  • Contributed core pieces of the technology that enabled SkillCorner to become a market leader in sports AI.
Technologies: Deep Learning, TensorBoard, TensorFlow, OpenCV, Python, Machine Learning, Artificial Intelligence (AI), Computer Vision, Image Recognition, Image Processing, NVIDIA CUDA, Convolutional Neural Networks (CNN), AI Programming, Linux, Amazon Web Services (AWS), Neural Networks, Data Science, Software Architecture, Algorithms, Bash, AWS DevOps, NumPy, GPU Computing, Cloud, Data Modeling, Artificial Intelligence (AI), Image Recognition, Machine Learning, Linux Servers, Object Detection

Research Assistant

2014 - 2015
Paris School of Economics
  • Served as an analyst on a project for Pôle Emploi, the French employment agency, to determine the effects of case-workers on placement rates of job seekers and evaluate the impact of using different levels of case-worker engagement.
  • Implemented on Stata, a statistical software used by economists, an approach for bounding the treatment effect in case of differential response rates to surveys on treatment and control groups.
  • Documented the data sources as well as the extraction, cleaning, and preparation pipeline we built during the project.
Technologies: LaTeX, STATA, SQL, SAS, Data Science, Data Modeling, Databases

Macro & Quant Analyst

2008 - 2012
Shânti Asset Management
  • Performed diverse roles, including research, pricing, risk management, performance analysis, investor communication, and in-house system development across the funds. Worked primarily in Mumbai with around two months per year spent at the Paris office.
  • Worked on a wide variety of asset types, including asset-backed securities, inflation-indexed bonds, inflation swaps, government debt, credit default swaps, futures, options, emerging equities, currencies, and Asian convertible debt.
  • Evaluated and monitored individual names in Asia ex-Japan on the basis of the strength of the issuer's cash flows and technical characteristics of the convertible bond. Built a binomial tree pricer for in-house valuation of CBs with embedded options.
  • Monitored the latest developments in economic activity, inflation, and monetary and fiscal policies and researched investment ideas in credit and FX in EM Asia, CEE, and Latin America. Produced monthly macroeconomics chart books.
  • Covered Indian macroeconomics as an input for sector and stock selection for the Indian equity fund. Authored the Economy and Society section of the monthly investor letter for the fund.
  • Valued cash-flow CLOs (collateralized loan obligations) through 2008-09 and developed an excellent understanding of CLO characteristics and market. Also familiar with synthetic ABS CDOs held in the portfolio.
Technologies: Bloomberg, Excel VBA, Quantitative Modeling, Time Series, Time Series Analysis

Computer Vision for Screening Mammography

Trained an object detection and segmentation model on a screening mammography dataset to detect breast lesions and classify them as benign or malignant. Using a deep model and various data augmentation strategies, we obtained high precision, recall, and ROC-AUC.

Telegram Chatbot Using an LLM and Voice Cloning

I created a Telegram chatbot using a large language model (LLM) similar to ChatGPT, along with voice cloning deployed on Google Compute Engine. The LLM we finally used was an uncensored open-source model that had to be primed with the right text prompt to suit our purposes.

Facial Keypoint Detection and Integration into an AR Application

I integrated a facial key-point detection system with some bespoke elements and a web front-end for a VR/AR (virtual reality/\augmented reality) application in the fashion industry. I used the detected key points to determine the correct placement of facial accessories irrespective of the angle of view.

Building Layout Parsing and Recognition

I developed a computer vision system for parsing building floor pans (images in jpeg, png, and more; PDFs; and CAD formats) and recognizing walls and rooms for downstream use in wireless network planning. The system used a pipeline of various image processing techniques, including line detection, contour detection, and area operations using GeoPandas.

Motion Detection Using Dynamic Background Subtraction

I built a motion detection system based on background detection that works even against dynamic backgrounds (e.g., moving water, tree leaves, and more) based on state-of-the-art algorithms in scientific literature.

Real-time Football and Player Detection

http://skillcorner.com/
Built a fast and accurate system for detecting players and the ball in broadcast football videos. I customized Faster R-CNN with many major and minor modifications to its architecture for this purpose. This work is now the core of a successful sports-tech startup.

Scene Text Recognition

Text detection and recognition in natural images for a client, using a pipeline of two deep neural networks and geometry-based post-processing. The 1st neural network detects the location of the text, while the 2nd recognizes the words.

Aspect-based Sentiment Analysis

I extracted aspect terms from reviews of places of interest (like restaurants) and predicted their associated polarities or sentiments. I used SpaCy, as well as a custom neural network: BiLSTM-CRT, for this project.

Recommender Systems

I built scalable recommender systems using PySpark, SparkML, and Hadoop. This includes popular public datasets such as MovieLens (small and full), as well as proprietary datasets furnished by clients for their projects.
2015 - 2017

Master's Degree in Computer Science

Université Pierre et Marie Curie - Paris, France

2012 - 2014

Master's Degree in Economics & Public Policy

Sciences Po - Paris, France

2004 - 2006

Master of Business Administration (MBA) in Finance

Indian Institute of Management Bangalore (IIMB) - Bangalore, India

2000 - 2004

Bachelor's Degree in Computer Science

National Institute of Technology - Surathkal, India

Libraries/APIs

PyTorch, NumPy, FFmpeg, SpaCy, Scikit-learn, Spark ML, Keras, TensorFlow, Google Maps SDK, REST APIs, Node.js, Pandas, OpenAPI, Python Asyncio, WebRTC, XGBoost, OpenCV, Socket.IO, Flask-RESTful, Vue, Natural Language Toolkit (NLTK)

Tools

ChatGPT, You Only Look Once (YOLO), Bitbucket, LaTeX, STATA, Weka, Sublime Text, TensorBoard, MATLAB, Celery, Named-entity Recognition (NER), CAD, AWS CLI, Bloomberg

Frameworks

Flask, GStreamer, Express.js, LlamaIndex

Languages

SQL, Python, Bash, C++, C, Java, JavaScript, Snowflake, SAS, TypeScript, Excel VBA

Paradigms

Data Science, DevOps, Back-end Architecture, Parallel Computing, Quantitative Research

Platforms

Docker, Google Cloud Platform (GCP), MacOS, Linux, Amazon Web Services (AWS), NVIDIA CUDA, Twilio, Raspberry Pi, Ubuntu

Storage

Databases, MySQL, Google Cloud, Redis, NoSQL

Industry Expertise

Project Management

Other

Artificial Intelligence (AI), Machine Learning, Deep Learning, Computer Vision, Natural Language Processing (NLP), Image Recognition, Image Processing, Generative Pre-trained Transformers (GPT), Chatbots, Machine Vision, Convolutional Neural Networks (CNN), AI Programming, Neural Networks, Algorithms, Data Analysis, GPU Computing, Training, Fine-tuning, Time Series, Time Series Analysis, Computer Vision Algorithms, Data Modeling, Artificial Intelligence (AI), Image Recognition, Machine Learning, Multithreading, Object Detection, Data Engineering, Video Processing, Full-stack, Reinforcement Learning, Facial Recognition, APIs, Corporate Finance, Generative Artificial Intelligence (GenAI), Large Language Models (LLMs), OpenAI, OpenAI GPT-3 API, Architecture, Integration, Speech to Text, Voice Recognition, Bots, Language Models, Audio, Software Architecture, Leadership, Hugging Face, SaaS, Prompt Engineering, Full-stack Development, Chatbot Conversation Design, Investment Banking, Quantitative Modeling, Finance, Team Leadership, Machine Learning Operations (MLOps), Back-end, Generative Pre-trained Transformer 3 (GPT-3), OpenAI GPT-4 API, Text to Speech (TTS), modal, BERT, Technical Architecture, Solution Architecture, LangChain, Pinecone, NLU, Applied Research, Transformer Models, Cloud Computing, Cloud, AI Chatbots, Project Management, Project Coordination, Linux Servers, Raster to Vector, FastAPI, PDF, Object Tracking, Generative AI, CTO, Deployment, Graph Theory, Investments, Accounts, Recurrent Neural Networks (RNNs), Long Short-term Memory (LSTM), Recommendation Systems, GeoPandas, Image Generation, AWS DevOps, LoRa, Medical Imaging

Collaboration That Works

How to Work with Toptal

Toptal matches you directly with global industry experts from our network in hours—not weeks or months.

1

Share your needs

Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2

Choose your talent

Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3

Start your risk-free talent trial

Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.

Top talent is in high demand.

Start hiring