Siddharth Sahani
ML EngineerAI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
AI is analyzing your overall score…
Identifying your key strengths…
Evaluating your skill match against the job requirements…
Assessing your cultural and operational fit
Principal Machine Learning Tech Lead | University Gold Medalist (2016)| IEEE Submissions Reviewer| ex-Kayzen, ex-Shipmnts, ex-IITM | Guest Speaker| AI/ML Mentor| 15k+ followers
AI/ML engineering leader with 10 years of experience building, shipping, and scaling production ML systems across adtech, logistics, legal-tech, and ed-tech. Currently leading ML strategy and cross-functional execution across three merged business units at a European DSP processing 50B+ daily bid requests under sub-15ms latency. Track record of building ML teams from the ground up (up to 15+ engineers), owning end-to-end product delivery from experimentation to production, and directly driving revenue outcomes. Deep hands-on expertise spanning real-time bidding, reinforcement learning, LLM/RAG systems, computer vision, and NLP paired with the business acumen to translate ML capabilities into measurable P&L impact.
SRM IST Chennai
Bachelor of Technology (B.Tech.), Computer Science and Engineering
January 1, 2012 – January 1, 2016
Omkarananda Sarawati Nilayam, Rishikesh
12th (Sr. Secondary), Science
January 1, 2011 – January 1, 2012
Omkarananda Sarawati Nilayam, Rishikesh
10th (Secondary)
January 1, 2009 – January 1, 2010
BidMachine
Principal Machine Learning Tech Lead
January 1, 2026 – Present
Remote
Appodeal, Inc.
Principal Machine Learning Tech Lead
January 1, 2025 – January 1, 2026
India · Remote
Kayzen
Senior Machine Learning Engineer
March 1, 2023 – January 1, 2025
Kayzen
Machine Learning Engineer II
March 1, 2021 – February 1, 2023
Great Learning
AI/ML Mentor
April 1, 2020 – March 1, 2021
Remote
Shipmnts
ML Lead & Product Owner
February 1, 2020 – February 1, 2021
Ahmedabad, Gujarat, India
Shipmnts
Senior Data Scientist & MLE
February 1, 2019 – February 1, 2020
Ahmedabad, Gujarat, India
Shipmnts
Data Scientist & MLE
February 1, 2018 – January 1, 2019
Ahmedabad, Gujarat, India
Shipmnts
Data Scientist
January 1, 2017 – January 1, 2018
Ahmedabad, Gujarat, India
InFoCusp
Machine Learning Engineer
June 1, 2016 – January 1, 2017
Greater Ahmedabad Area
Zaya Learning Labs
Machine Learning Researcher
December 1, 2015 – May 1, 2016
Mumbai Metropolitan Region
Indian Institute of Technology, Madras
Undergraduate Research Fellow
May 1, 2015 – May 1, 2016
Greater Chennai Area
SRM TEAM ROBOCON
Core Team Member
March 1, 2014 – August 1, 2014
Greater Chennai Area
SRM TEAM ROBOCON
Autonomous Robot Programmer
March 1, 2013 – February 1, 2014
Greater Chennai Area
Document Layout Drift Detector
January 1, 2021 – October 1, 2021
- Identifying page layouts deviating from a fixed set of 20 Shipping Documents which have more tablular and singular fields vs documents that just have drawings, or logos, or signatures - Continual Monitoring through Airflow job and firing alerts when such pages are beyond a threshold - Auto curate a dataset of drifted values that can be further verified by a team of annotators to tag as a new category (via Semi supervised feedback)
Wheeze & Crackle Detection
June 1, 2017 – July 1, 2017
The most used signal processing techniques to produce features for classification are based on spectral analysis, wavelet transforms and statistics. Mel Spectrogram represents an acoustic time-frequency representation of a sound. It is the result of transforming a spectrogram’s values into the mel scale. While reading more about state-of-art MFCC Mel-frequency Cepstral Coefficients, and why CNN architecture makes it inefficient to use because of the inherent parameter sharing to decrease the number of parameters needed for high dimensional input grids. A Convolution Network always has the same number of parameters, even with bigger or smaller sized input grids. MFCC features by nature are not robust to shifting. If the preprocessing steps are not robust and are sensitive to internal/external noise and quality of records, SVMs, kNNs or even CNNs for that matter may perform not so well on shifted lung sounds owing to the variety of stethoscopes What more can be experimented Hyperparameter tuning using Auto-ML or Amazon Sagemaker in Convolution architecture. I would also want to study the effect of dropout in different architectures. A stateful recurrent model is one for which the internal states (memories) obtained after processing a batch of samples are reused as initial states for the samples of the next batch. This allows processing longer sequences while keeping computational complexity manageable. RNNs that can tap the power of MFCCs as Feedforward networks have access to all input features without the utilization of shared parameters, combined with the temporal context of the data, making it a much better architecture for interpreting MFCC input. As per another recent study, VGGish network can better handle audio data, and RNN can better process time series data. From the view of the spatial domain and the time domain, a lung sound recognition model based on VGGish-BiGRU combines both convolutional and recurrent properties.
Badminton playing Robots & Playfield Child-Parent robots
May 1, 2013 – September 1, 2014
Programmer in SRM Tech Team Robocon • Coded 3-wheeled, 2-wheeled and even 1-wheeled robot with PID for line following. • Worked with Ultrasonic, IR, Hall-effect sensors, Magnum and industrial grade Maxon motor, IMU, Polulu, Hercules, and Sabertooth motor drivers; and programmed them as per needs of our play. • Was In-Charge for the Autonomous Child robot in ABU Asia-Pacific Robocon-2014. Our team bagged “Best Economical Robot”, and settled with 16th rank. • Trained juniors for the next competition, for about 2 months; and moved on to study the inculcation of actual Artificial Intelligence in robots.
Introduction to Cloud Computing (Score: 100%)
edX
June 24, 2026 – Present
Responsible AI: Applying AI Principles with Google Cloud
Google Cloud Skills Boost
June 24, 2026 – Present
Introduction to Responsible AI
Google Cloud Skills Boost
June 24, 2026 – Present
Introduction to Generative AI
Google Cloud Skills Boost
June 24, 2026 – Present
Introduction to Large Language Models
Google Cloud Skills Boost
June 24, 2026 – Present
Generative AI Fundamentals
Google Cloud Skills Boost
June 24, 2026 – Present
Cultural Fit Analysis
The candidate's diverse project portfolio, ranging from robotics to Document AI and real-time bidding systems, demonstrates adaptability and a broad interest in applying ML across different domains. Their progression through various roles at Shipmnts and Kayzen, taking on increasing responsibilities, indicates a growth mindset and commitment. The mentoring role at Great Learning also suggests a willingness to contribute to the community and develop others. The experience aligns well with a dynamic, innovation-driven environment.
Soft Skills & Operational Fit
The candidate's experience as a Principal ML Tech Lead and Senior ML Engineer, leading teams and acting as an integration layer between various departments, suggests strong leadership, communication, and collaboration skills. Their ability to stabilize real-time systems under high load and optimize complex pipelines indicates excellent problem-solving and operational management capabilities. The mentoring experience also points to strong interpersonal and knowledge-sharing skills.