Resume 💼
EDUCATION
Master of Science in Artificial Intelligence
Coursework: Machine Learning, Computer Vision, Natural Language Processing, Deep Learning, FACT-AI, Reinforcement Learning, Information Retrieval, Recommender Systems, AI for Medical Imaging
Bachelor of Technology in Information Technology
Coursework: Data Structures, Data Science, Advance ML Techniques, Probability and Statistics, Calculus, Linear Algebra, Modelling and Simulation
WORK EXPERIENCE AND INTERNSHIPS
TrendWatching – Amsterdam, Netherlands
- Architected an end-to-end pipeline integrating web crawlers for news sourcing and automated curation, designed to scale the innovation database capacity by 5x, enabling faster trend discovery and broader coverage.
- Created a recommendation engine to surface contextually similar innovative articles from large-scale raw news feeds, enhancing content relevance and enabling early identification of emerging trends.
Outplay (Series A, acquired by JungleWorks) – Hyderabad, India
- Designed and implemented a Conversation Intelligence pipeline to transcribe and analyze over 10,000 sales calls, integrating models for Tone Detection, Speaker Identification, and Summarization, achieving 95% accuracy across tasks.
- Trained and deployed an in-house Automatic Speech Recognition (ASR) system for real-time and offline transcriptions, with models achieving a word error rate of 25% or lower.
- Fine-tuned open-source Text-to-Speech models for voice cloning, creating natural-sounding voices for personalization and dialer phone call verification.
- Automated scaling of AWS EC2 instances, reducing server idle time and cutting monthly EC2 costs by 90%.
- Built “Live Call Assistant” using WebSockets, Kafka, and deep learning models to extract insights and trigger real-time alerts such as question-answering, speaking rate, sales pitch analysis, and ice breaker suggestions.
- Developed a Named Entity Recognition (NER) system to extract key information from Out-of-Office emails, including return dates, prospect names, and alternate contact details, with 90% accuracy.
- Constructed a novel dataset and applied prompt engineering techniques to fine-tune OpenAI's GPT-3.5 to write emails, significantly reducing response times and improving outreach efficiency.
Cornerstone Solutions Pvt. Ltd. – Bangalore, India
- Collaborated with the Robot Operating System (ROS) team to aid underwater bot navigation within the Gazebo simulation, finding valuable insights regarding the conditions of underwater tunnels.
- Automated underwater bot navigation by training reinforcement learning models that incorporated real-world bathymetry data and mitigated water disturbances using simulated sonar sensors.
Intentico Inc. (acquired by Movius) – Bangalore, India
- Optimized client Chatbots with AI using Long Short-Term Memory Networks (LSTM), streamlining client processes, and improving user engagement.
- Developed sentiment analysis models for Hindi and Arabic using CNN architectures and implemented automated test scripts with Selenium for performance assessment.
Lynkit Pvt. Ltd. – Delhi, India
- Performed comprehensive data analysis on a vast dataset of real-time GPS device track data, comprising over 100,000 rows, and generated visualizations to extract insights.
- Developed Regression and Classification models to uncover tamper alerts in GPS-enabled RFID shipment container locks.
SKILLS AND COMPETENCIES
Programming Languages & Tools: Python, JupyterLab, Git/Github, Jira, Grafana, SonarQube, Datadog, Sentry
Databases: PostgreSQL, MongoDB, MySQL
Frameworks: PyTorch, Tensorflow, Keras, Jax, Kafka, FastAPI, Streamlit, Gradio
MLOps: Docker, MLflow, Kubeflow, Weights & Biases, Jenkins Cloud Platforms: Google Cloud Platform (GCP), Amazon Web Services (AWS)
PUBLICATIONS
Thakur, U.*, Khan, M. H.*, Changlani, M., Dotsinski, A., Mahadevan, A. K., Kechagias, I., Najdenkoska, I. CLaRE: CLIP with Latent Reconstruction Errors for Generated Face Detection. In Proceedings of the 2025 Deepfake, Deception, and Disinformation Security Workshop (3D-Sec ‘25), October 13–17, 2025, Taipei, Taiwan. ACM, New York, NY, USA 11 Pages. (* Equal Contribution)
Dotsinski, A.*, Thakur, U.*, Ivanov, M., Khan, M. H., & Heuss, M. On the Generalizability of “Competition of Mechanisms: Tracing How Language Models Handle Facts and Counterfactuals”. Transactions on Machine Learning Research (TMLR), 2025. (* Equal Contribution)
Yadav, A., Thakur, U., Saxena, R., Pal, V., Bhateja, V., & Lin, J. AFD-Net: Apple Foliar Disease multi classification using deep learning on plant pathology dataset. Plant and Soil, 477, 1–17, 2022.