Advertisement Number: P.Rect./R&D/2023/112

About the project: The continuous ambient air quality monitoring stations (CAAQMS) that are used by government agencies to monitor the concentration levels of PM are expensive.Low-cost sensors have emerged as a potentiallycost-effective option for dense air pollution monitoring networks. But they require in-field calibration to improve their precision and accuracy in comparison with CAAQMS. We use machine learning (ML) based regression techniques to calibrate the LCS readings.

Our recent publications involve domain adaptation for sensor calibration. (S. K. Jha, et al. (2021): Domain adaptation based deep calibration of low-cost PM2.5 sensors) applies simple fine tuning based domain adaptation technique for calibration. (K. Yadav, et al. "Few-shot calibration of low-cost air pollution (PM2. 5) sensors using meta-learning) applies model-agnostic meta learning technique for few shot domain adaptation based calibration.

About the group/team: This is a highly interdisciplinary project with team members from the Electrical, Civil and Computer Engineering disciplines. The team has a high degree of specialization in air quality monitoring as well as machine learning/AI.

About our machine learning/AI work: The team, led by Prof. Tripathi and Prof.Arora, engages in cutting-edge research in all major areas of machine learning and AI. The team has a strong tradition of student-led research and provides a supportive and motivating environment to all its members.

About our air-quality monitoring work: The team has made impactful contributions at the national scale to address the challenges of Air Pollution and Climate Change. Under the leadership of Prof. Tripathi, the team has built ground-breaking innovative approaches for indigenously built sensor-based network technologies for nation-wide urban air quality monitoring and Real Time Source Apportionment (RTSA). Recently, the group has been awarded the International Centre of Excellence on advanced pollution monitoring technologies recognized by the Office of the Principal Scientific Advisor to the Government of India. This work will be carried out as a part of this Centre. RTSA work has been done in Delhi with support from CPCB and is currently being carried out with support from Swiss Development Corporation in the cities of Lucknow and Pune.The group has also established sensor networks in 5 cities (Jaipur, Chennai, Guwahati, Kanyakumari and Delhi) with support from Ericsson.

About the position

We are looking for One Project Post-Doc who would be in-charge of data analysis and building machine learning models for sensor calibration and converting SAAQM data to SA insights.

Responsibilities: The candidate would be responsible for taking data generated by multiple sensors and regulatory grade equipment, and using data analysis and ML/AI techniques for calibration and to generate insights into the data such as outliers, trends, latent correlations. The candidate will be involved in research on developing new machine learning models that can do the calibration efficiently.

Relevant skills & experience

• Strong background in machine learning, data science and artificial intelligence with fluency in basic ML models such as regression, classification, and clustering.
• Strong background in mathematics, especially probability theory and time series analysis
• High ethical standards when performing data analysis and reporting results.
• Willingness to work in an interdisciplinary setting and collaborate closely with team members from other fields, learn about new technologies used in those domains and incorporate this domain knowledge to create more powerful machine learning models.
• Fluency in Python coding and use of ML libraries such as numpy, scipy, and sklearn.
• Prior experience with deep learning libraries such as Keras, PyTorchortensorflow.
• Significant background in handling and processing large data sets.
• Impressive publication record in high-impact ML/AI conferences and/or journals.
• Familiarity with advanced topics such as time series models (VAR, VARIMA and deep learning variants such as LSTM, GRU, Transformers) and Gaussian processes will be a plus but is not essential.
• Past background in interdisciplinary research and working with raw data e.g., from instruments or sensors will be a plus but is not essential.

Compensation & benefits

The candidate will get the opportunity to work in a highly motivating and supportive environment. The candidate will be able to avail hostel/family accommodation facilities at IIT Kanpur (according to availability). A dequate support for travel and living costs will be provided (per diem etc.) in case travel or field trips are required. With regards to the medical facilities, only OPD facility for self shall be available at the Health Centre of IIT Kanpur without any reimbursement of cost of medicines. Any outside referral shall not be reimbursed. For the purpose, a medical booklet shall be provided to the candidate on request.

Essential Qualification

• Ph.D. in Computer Science /Electrical Engineering or closely related fields with specialization in Machine Learning/ Artificial Intelligence.

Duration of Appointment 

• The post is purely temporary and is on contractual basis.
• The appointment for the post is up to two years. Appointment will be initially for one year. Renewal of appointment will be subjected to the performance of the candidate.
• Salary: Rs. 60,000/65,000/70,000 per month (as per experience)

The departments reserve the right to fix suitable criteria for short listing of eligible candidates satisfying qualifications and experience. The selection will be based on interviews: online or in person. Short listed candidates will be informed via email about the date of interview. 

Interested candidates may apply via email (to dstccp@gmail.com) giving full details qualifications, experience and two recommendation letters with copies of relevant certificates by May 25, 2023. *Extended till June 03, 2023.

Please mention the post applied for and advertisement no. in the subject line of email.