Job Title

Company Description is a forward-thinking technology company focused on delivering cutting-edge solutions using artificial intelligence and machine learning. Based in Chandigarh, is dedicated to addressing complex challenges by leveraging advanced research, data-driven insights, and a strong commitment to innovation. By fostering collaboration and continuous learning within a highly skilled team, aims to create impactful technologies for businesses and industries. Join a company that is shaping the future of technology with intelligence and creativity. About The Role You will research, develop, and deploy multiple ML models that run at the heart of 's intelligence platform. The data you work with is real, high-frequency, and physically meaningful — 3-phase power signals, soil moisture at three depths, vibration spectra from pump motors, groundwater levels, satellite NDVI. The models you build will advise 50,000+ farmers on when to irrigate, detect bearing failures three weeks before they happen, schedule feeder-level demand response, estimate water savings for carbon credit verification, and predict aquifer depletion across entire districts. Your work goes from notebook to production and then to a farmer's WhatsApp message. Models You Will Build Irrigation advisory model — predict daily crop water requirement using soil moisture, weather forecast, crop type, and growth stage; output must be explainable to a farmer in two sentences Pump bearing failure prediction — classify fault type and estimate remaining useful life from vibration FFT and motor current signature; 21-day advance warning is the target Sensorless water flow estimation — infer flow rate from 3-phase motor current signature analysis (MCSA) without a mechanical flow meter; this is novel IP and is patent-pending Feeder demand response scheduler — optimise pump scheduling across a group of farms on a shared electrical feeder to reduce peak load and minimise irrigation delay simultaneously Groundwater depletion model — forecast aquifer drawdown rate from IoT extraction data, CGWB borewell levels, satellite gravity anomaly, and historical monsoon patterns Water savings verifier — calculate and certify energy and water savings per enrolled farm for carbon credit MRV (Monitoring, Reporting, Verification); output must satisfy Verra registry audit standards Water quality anomaly detector — identify contamination events and sensor drift from multi-parameter water quality time-series (pH, EC, turbidity, nitrate, fluoride) Foundation model contribution — as fleet data grows, contribute to a cross-farm, cross-state agricultural intelligence foundation model trained on 's proprietary sensor dataset What You Will Do Own the full ML lifecycle for your models — data exploration, feature engineering, model selection, training, validation, and deployment — not just the modelling step Design and implement physics-informed constraints where applicable — the irrigation model uses FAO-56 Penman-Monteith as a hard boundary; the aquifer model uses Theis equation constraints — your ML must respect physical laws Build and maintain the feature engineering pipeline on TimescaleDB: time-series features, rolling statistics, spectral features from FFT, satellite-derived NDVI, and weather-derived ETo Write model cards and evaluation reports for every model that goes to production — accuracy, failure modes, known biases, and retraining triggers all documented Collaborate with the firmware team on edge deployment — the bearing failure and irrigation models run on-device on constrained hardware; you quantise and optimise for INT8 inference Implement SHAP-based explanation pipelines for every model — every prediction must generate a plain-language reason that a farmer or government official can understand Monitor deployed models in production — data drift, prediction drift, performance degradation — and own the retraining pipeline when monitoring alerts fire Contribute to patent filings — several model are novel and patentable; you will author Invention Disclosure Forms for innovations you develop  What We Are Looking For Required 3–7 years building and deploying ML models that are live in production serving real users or systems, not only research or academic work Strong time-series ML: LSTM, Transformer, CNN on sensor data; feature engineering from raw signals; multi-step forecasting Solid Python ML stack: PyTorch or TensorFlow, scikit-learn, Pandas, NumPy, SciPy — writes production-quality code, not just notebooks Hands-on experiment tracking and model registry discipline — uses MLflow or equivalent; knows the difference between a model version and a model deployment Experience deploying models to production: containerised serving, latency SLAs, rollback capability B.Tech / M.Tech / MS in Computer Science, Electrical Engineering, or a quantitative field Good to Have Signal processing: FFT, filter design, spectral feature extraction from vibration or power signals — directly applicable to MCSA and bearing fault detection Physics-informed neural networks (PINNs) or physics-constrained modelling — any domain where physical equations constrain the model output Reinforcement learning for scheduling or control problems — directly applicable to the demand response scheduler Edge ML deployment: TFLite Micro, ONNX Runtime, INT8 quantisation for constrained embedded hardware Agricultural, environmental, or energy domain knowledge — familiarity with evapotranspiration, groundwater hydrology, or power systems Satellite remote sensing: Sentinel-2 processing, NDVI derivation, Google Earth Engine scripting