From apes to wolves, we are changing the face of conservation tech: Rushikesh Chavan, Director, The Habitats Trust | Technology News

Rushikesh spoke to indianexpress.com on the challenges of using tech for conservation, the impact of their grants, the new tech they are working on, and how AI will impact the conservation tech space. Edited excerpts:

Venkatesh Kannaiah: Can you give us an overview of the work of The Habitats Trust?

Rushikesh Chavan: Founded in 2018 by Roshni Nadar Malhotra of HCL Technologies, THT is focused on challenges arising from biodiversity loss, the climate crisis and global water scarcity.

We work in two ways. We provide grants to partners ranging from Rs 3 lakh to Rs 1 crore to support their efforts. Beyond grants, we also identify specific themes where we fund partners and are part of the execution process.

When we find that no organisation is working on an important issue in a geography, we execute the project ourselves.

We also work with various government organisations. One among them is MITRA, the policy think tank in Maharashtra that works on the lines of NITI Aayog, integrating ecological priorities into their regional planning efforts, such as the Vidarbha plan.

Venkatesh Kannaiah: Tell us about some of your grantees/partners and how they are using tech to tackle issues of conservation and biodiversity?

Rushikesh Chavan: Hoolock Gibbons are elusive primates, and the only ape found in rainforest habitats of Northeast India. Given the difficulty in direct sightings, scientists are using its vocalisation habits to estimate its population.

THT, in collaboration with Conservation Initiatives, is using acoustic recorders to detect these gibbons. We are also creating AI and machine-learning-based models to process acoustic data collected by these recorders and identify gibbon calls and their numbers in a given locality.

The Indian wolf is a species under threat, and we are working with The Grasslands Trust to monitor wolf packs and their behaviour using satellite radio collars, thermal drones and camera trapping. While radio collars help track the daily movement of wolves, thermal cameras help in detecting their behaviour at night.

Tracking land use and land cover using satellite imagery is critical for conserving wildlife and biodiversity. Open Natural Ecosystems (ONEs) are one of the largest types of habitat forms in India that include grasslands, scrublands and open savannah. These habitats have been incorrectly categorised as wastelands in government atlases. We are working with pioneering researchers like MD Madhusudan and Pradeep Koulgi to accurately map these habitats and thereby restrict their overexploitation.

Venkatesh Kannaiah: Tell us about your Tech for Conservation programme and its highlights.

Rushikesh Chavan: At The Habitats Trust, we see tech as a tool and not as a solution in itself. In India, technology only works if the surrounding ecosystem supports it. So in our Tech for Conservation programme, we focus on three themes.

First, once we identify a critical conservation gap, we develop tech where necessary. For example, our work in bioacoustics has been very effective. We have also developed AI-based camera trap models to help sort and process large volumes of wildlife data. We have built our own ROVs (Remotely Operated Vehicles) to study marine and ocean ecosystems, sending them down to depths of nearly 200 metres. This kind of work is rare in India.

Second, we assess whether the ecosystem exists for that technology to be effective. We partner with researchers, conservationists, and field organisations to deploy the tech.

Conservation tech is expensive. Since we are a philanthropic organisation, we invest in building the tech and then make it open access wherever possible.

Venkatesh Kannaiah: How are you using AI and machine learning in your conservation efforts?

Rushikesh Chavan: First, at a systems level. For example, in Maharashtra, we are working with NITI Aayog, using AI, GIS, and large-scale data modelling to design growth pathways that integrate ecology and the economy. If you don’t maintain the ecological integrity of a landscape, there is no point focusing only on a single species.

Second, we use AI and machine learning primarily for generating evidence. We use evidence actively to inform decisions, shape policy, and design interventions.

For example, all our camera trap data processing uses AI/ML models. Our bioacoustic work relies heavily on machine learning to separate species calls from ambient noise. Video and sonar analysis from marine explorations also uses ML models. Today, if you are not using AI and ML, you are losing out.

We have also explored large language models (LLMs). But that space is evolving very rapidly with various models, so we are constantly rethinking our strategy there.

So broadly, AI, ML, and LLMs are used by us for large-scale data analysis, for pattern recognition in ecological datasets, systems modelling, scenario building and decision-support frameworks.

Where we have not fully cracked the problem yet is implementation. For instance, imagine a drone with infrared flying over a forest. AI can detect hotspots or anomalies, say, early-stage forest fires. That gives us information. But how do you convert that into something a forest officer can immediately act upon, cost-effectively? That is still a gap.

In India, cost is a serious constraint. Advanced AI systems can be expensive to deploy at scale. AI in India still has to cross that cost-effectiveness barrier for large-scale, on-ground operational deployment.

AI is extremely powerful for knowledge generation, evidence building, and strategy design. But when it comes to direct intervention, replacing or automating ground-level work, we are not fully there yet.

Venkatesh Kannaiah: What tools have you built for helping conservation efforts?

Rushikesh Chavan: We have developed an underwater ROV to explore deeper marine ecosystems. It allows us to capture video, imagery, and side-scan sonar data at depths of up to 200 metres; areas that are otherwise very difficult to study in India.

The second innovation is a multi-array microphone system. Normally, if you place a microphone in a forest, it captures sound from a single direction. We developed a multi-directional microphone array that can not only record sounds but also triangulate where exactly the sound is coming from. That spatial accuracy is critical for ecological studies.

Now, coming to software solutions. One major tool we’ve developed is a camera trap image-sorting platform. India generates millions of wildlife photographs through camera traps. We built a base AI model that allows users to upload their camera trap data and automatically sort images.

On top of that, there are plugins for species classifiers that can be added. For example, if a forest department wants to isolate tiger images, a plugin can extract all tiger photographs from the dataset. We’ve provided this tool for free to places like Kaziranga Tiger Reserve, especially during large exercises like the All India Tiger Estimation, where they generate lakhs of images with limited capacity to process them.

The next layer of software relates to analysing video and sonar data, especially from the rover. Camera traps produce still images, which are easier to process. But underwater video and side-scan sonar data are far more complex. So we are developing models that can analyse video streams scientifically, not just create a species list, but apply statistical models to generate robust ecological insights.

Then there is our bioacoustics work. Using the multi-array microphones, we capture large volumes of forest sound data. We talked earlier about the Hoolock Gibbon project. The next stage is identifying individuals through vocal signatures. Just as you and I sound different, animals have unique vocal patterns. There is a concept in acoustic science that may allow differentiation between individuals. If that works for gibbons, and when combined with directional data from the microphone arrays, we can apply statistical models to reliably estimate the number of animals or birds that exist in a given landscape. That would be a breakthrough.

We have developed around five to six major technology solutions, some of which are in the public domain and actively being used, and others are at various stages of development. As and when the models are validated, we release them openly.

Venkatesh Kannaiah: Tell us about five innovations or technologies that are changing the face of conservation.

Rushikesh Chavan: There is the Acoustic Microphone Array tech for tracking species that are more heard than seen, which we spoke about.

There is the ROV tech. Humans can safely dive and study species only at depths of around 30-50 metres, but these underwater robots, unoccupied and piloted from the surface, become crucial for deepwater tasks, scientific research, and surveillance.

There is the metabarcoding and genetic analysis tech. DNA analysis only allowed us to know if a species was present at a place, only if tissue, hair or excreta is available. Now, advanced techniques of DNA analysis can tell us if a species is experiencing a population decline, suffering from diseases or vulnerable to change.

There is thermal imaging to track and monitor elusive nocturnal species. Nocturnal behaviour monitoring has been a challenge in situations of human-wildlife conflict. Now, thermal-imaging-equipped drones are used for surveillance of sites where a conflict has happened and have often helped managers to safely locate and rescue animals and alert people.

There is auto-classification and GSM-enabled camera trap tech. Camera traps are deployed to record wildlife, and now advancements in models are helping us identify species quickly and send that information directly to our phones and systems. It helps security agencies detect poachers.

Venkatesh Kannaiah: Tell us about global out-of-the-box startups working in conservation tech and the problems they are solving.

Rushikesh Chavan: There is Conservation X Labs, which uses AI, advanced sensors and open innovation to help conservationists respond to threats such as poaching, invasive species and habitat loss.

There is Nightjar Tech, which develops AI-powered wildlife monitoring hardware and an intelligent camera-alert system that detects animals, people and threats in real time.

Sara Beery Lab works on computer vision and machine learning for ecological monitoring, creating AI methods to automate wildlife image and sensor data analysis for conservation science.RoboticsCats is a Hong Kong-based machine vision and AI startup that builds environmental monitoring and early wildfire detection systems using computer vision to detect smoke, fire and biodiversity cues from surveillance cameras.

There is also EarthRanger, a data-driven software platform widely used by conservation teams to integrate field data, sensor feeds and AI insights into real-time decision-making tools that improve wildlife monitoring, conflict mitigation and habitat protection globally.

Venkatesh Kannaiah: What do you think are one or two low-hanging fruit, where AI can intervene in this tech conservation space?

Rushikesh Chavan: One low-hanging fruit for AI and ML is pollution monitoring and remediation.

Using satellite imagery and remote sensing, AI models can estimate various parameters for pH or contamination levels. Once you have that intelligence layer, the next step is automation.

For example, invasive species like water hyacinth choke many Indian water bodies. Removing them manually at scale is extremely difficult. AI-driven monitoring systems combined with small, automated, solar-powered removal units could identify hotspots and deploy targeted mechanical action. That is a space where AI plus automation can make a difference.

If we think about something like cleaning the Ganga, a purely human-driven effort at that scale is nearly impossible. But AI-assisted monitoring and mechanised interventions could significantly improve efficiency.

A second opportunity is decision-support systems for managers. AI-powered dashboards can integrate camera traps, satellite imagery, fire alerts, and ground reports into a single interface.

A third promising space is ecological restoration. India has immense biodiversity, and restoration requires local ecological knowledge — what species to plant, in which soil, under what rainfall regime. That expertise currently sits with a handful of specialists. AI models can integrate biodiversity data, soil profiles, rainfall data, and historical vegetation patterns to recommend species mixes.

Venkatesh Kannaiah: How is the tech ecosystem around conservation and biodiversity in India?

Rushikesh Chavan: One major challenge is cost. Conservation tech is expensive, and unlike sectors like finance or e-commerce, conservation doesn’t have deep-pocketed customers.

The second challenge is scale and sustainability for startups. Many startups build excellent proof-of-concept tools but struggle to make the business model viable.

Another is dependency on large technology platforms. Conservation tech developers are often adapting existing tools — satellite imagery platforms, cloud infrastructure, mapping tools — rather than building everything from scratch. If a company changes its pricing model, API access rules, or licensing terms, that disruption directly affects conservation applications.

Venkatesh Kannaiah: Tell us about the two big tech challenges you want to solve.

Rushikesh Chavan: The challenge is how to design technology that is usable in real-world Indian conditions. Tools that are lightweight, low-bandwidth, device-agnostic, and easy to deploy.

The second challenge is local adaptation. What works in Chennai may not work in Bengaluru or Delhi. Ecological systems differ. Soundscapes differ. Landscapes differ. Even governance systems differ. A restoration model built for Himalayan ecosystems won’t automatically apply to dry grasslands.

Otherwise, you end up building a new technology for every single use case, which is inefficient. Development takes time, effort, and money. And by the time you finish building something, the broader technology landscape may have already shifted, making parts of your system obsolete.