Souvik Mandal
Menu
Portrait of Souvik Mandal

Researcher · Educator · Data Scientist · Data engineer

I build scalable, trustworthy data and AI solutions to real-world problems.

I am an academic researcher turned technologist, focused on making education ready for a rapidly changing world by connecting learners, teachers, and technology. I hold a PhD in quantitative behavioral sciences, spent 5+ years in research and teaching at Harvard, and led data engineering and AI architecture for 2 years at an Ed-Tech platform with 65M+ global users.

See my work

Mission

Equitable access and opportunity for…

Quality Education

through which learners become thinkers, explorers and problem-solvers.

Health

connecting health requirements with researchers, providers, insurers, pharma and public agencies.

Climate Resilience

protecting ecosystems at risk, connecting communities, researchers, conservation groups, governments and industry.

…through data and AI.

My work

From modeling behavior to shipping AI products

During my PhD, I worked on statistical modeling of searching, navigation, and cooperative behavior. During my postdoctoral research at Harvard, I built data and ML pipelines that automated extraction and analysis of behavioral data from videos at large scale. At LabXchange, an online education platform, I took data and AI pipelines into production. Now I build data and AI solutions across the full lifecycle, from problem-framing to deploying the system.

My experience spans:

Education Technology

That serves learners, educators, and policymakers, combining behavioral science, data science, and AI/ML.

Teaching & Training

Teaching research methods and problem-solving using data and AI/ML. Training educators on effective use of AI.

Research

Focused on learning, decision-making, and cooperation using data science, computer vision, and other AI/ML workflows.

AI-native Startups

Building internal knowledge-enabled, multi-agent, and on-device AI products, now in use by alpha users.

Education Technology

Interest-driven, interdisciplinary learning for real-world problem-solving

I am a proponent of learning through solving problems and answering questions that learners are interested in. This approach connects knowledge and skills across disciplines, sustains engagement and develops transferable skills, judgment, and agency to tackle unfamiliar problems. This adaptability to novel contexts is especially crucial in this rapidly changing world.

While technology has expanded access to educational content, learning experiences too often remain fragmented into disconnected subjects, with limited opportunities to apply the learning to the interest of the learners. AI grounded in carefully curated educational resources offers an opportunity to organize learning around individual interests and practical problem-solving, connecting foundational knowledge with inquiry and critical thinking. Combining my expertise in behavioral science, pedagogy, data, and AI/ML, I build tools across the education ecosystem - for learners, educators, and policymakers.

Experience in Ed-Tech

  • Data Engineer & AI Solutions Architect, LabXchange, Harvard University · 2023–2025

Relevant skills

  • Cognitive Science
  • Curriculum & Instructional Design
  • Interdisciplinary Learning
  • Data Infrastructure & Analytics
  • Generative AI
  • Educator training

Teaching & Training

Learning through solving "your" problems and finding answers to "your" questions

Pose estimation turned into a fencing analytics dashboard, an example from the course. Shown in slow motion.

I bring my learning philosophy to my teaching in Computational Behavioral Sciences, a course I developed and lead at Harvard University. In this course, students start from their own question about behavior, and learn the research methods, computational thinking, data science and applied AI/ML to answer it, building domain knowledge along the way. All the course matrials including codes are freely available on GitHub.

With information and generative AI at everyone’s fingertips, educators can no longer serve just as sources of information. They must teach problem-solving: constructing logic, collaborating, and finding, evaluating and assimilating information. I co-designed and delivered training for about 8,000 educators in India through LabXchange, and I regularly lead workshops on the changing role of educators in the age of AI. I plan to take this hands-on approach to collaborative learning, with teams mixed across countries rather than competing as national blocks.

Experience in Teaching & Training

Relevant skills

  • Instructional design
  • Research methods
  • Computer vision
  • Responsible AI
  • Open educational resources
  • Educator training

Research

Understanding behavior across species and scales

Automated pose tracking of ants.

Behavior is where biology, experience, and environment meet. I studied how individuals and collectives forage, navigate, remember, learn, make decisions, and cooperate, across wasps, ants, honey bees, and robots.

My approach is quantitative. I pair field and laboratory experiments with computer vision, then analyze the results using statistical modeling, causal inference, and machine learning, including deep learning. I developed Ethoscope, a Python pipeline for automated, large-scale extraction of behavior from video, freely available on GitHub for academic use.

At Harvard, I led research teams, line-managing and mentoring 40+ students and project staff, from hiring and performance reviews to tracking work against deadlines and budgets.

Future research interests

  • Large-scale behavioral science of individual, collective, and cooperative behavior in humans and other social animals
  • Cognition and learning in the context of education
  • Behavior in personal and public health

Experience in Academic Research

Data Science Project

  • How credible are corporate climate pledges? · 2026. A credibility score for the top 50 U.S. power-sector parent companies, fusing SEC 10-K filings and Net Zero Tracker pledges (scored with a multi-LLM pipeline), EPA GHGRP and CEMS emissions data, and Climate TRACE satellite estimates, explored through an interactive dashboard.

Relevant skills

  • Experimental design
  • Statistical modeling
  • Bayesian statistics
  • Causal inference
  • Time-series modeling
  • Computer vision
  • Deep learning
  • HPC
  • Python
  • Research team leadership
  • Line management
  • Data fusion
  • LLM pipelines
  • Remote sensing data
  • Dashboards

Watch a museum video about my research →

Selected publications
  1. Ganga Prasath S*, Mandal S*, Giardina F*, et al. (2022). Dynamics of cooperative excavation in ant and robot collectives. eLife
  2. Mandal S, Brahma A (2019). Getting older, getting smarter: ontogeny of the foraging behaviour in the tropical social wasp Ropalidia marginata. Journal of Experimental Biology
  3. Mandal S (2018). How do animals find their way back home? A brief overview of homing behavior with special reference to social Hymenoptera. Insectes Sociaux
  4. Nouvian M, Mandal S, et al. (2018). Cooperative defence operates by social modulation of biogenic amine levels in the honey bee brain. Proceedings of the Royal Society B
  5. Brahma A, Mandal S, Gadagkar R (2018). Emergence of cooperation and division of labor in the primitively eusocial wasp Ropalidia marginata. PNAS

Full list on Google Scholar. * joint first authors.

AI-native Startups

Trustworthy AI products, built end to end

AI products that handle personal data, like a career history or a live camera feed, have to earn trust by design, not by promise. I am building two, with co-engineers, both in alpha testing with invited users.

Founder · 2025–present

CViable

A multi-agent résumé, cover-letter, and mock interview studio, grounded in the user’s career record and tailored to individual jobs. An independent agent verifies every factual claim against its source before export. Users can inspect and edit every prompt steering the system.

  • Multi-agent systems
  • Hybrid RAG
  • LLM evaluation
  • Python
  • Django
  • AWS

Yogi’s Bliss

A vision-enabled, voice-first conversational AI yoga guide. Real-time pose estimation runs entirely in the browser, so user’s videos never leave their device. An on-device voice loop lets people practice without watching the screen. Safety rules are enforced in code rather than left to the language model.

  • Computer vision
  • On-device ML
  • Voice AI
  • Privacy engineering
  • WebGPU
  • Progressive web apps

Transferable skills

Skills built across science, education, and AI

Research and Scientific Inquiry

Turning open questions into testable designs and published evidence.

  • Research design
  • Experimental design
  • Quantitative behavioral science
  • Scientific writing
  • Technical writing
  • Peer review
  • User Analytics

Data Science and Statistics

Analysis that support scientific findings and business decisions.

  • Statistical modeling
  • Data preparation
  • Data mining
  • Causal inference
  • A/B testing
  • Survival analysis
  • Longitudinal analysis
  • Python
  • R
  • SQL
  • Data visualization

Machine Learning

Predictive models that are measured, explained, and deployed.

  • Supervised learning
  • Unsupervised learning
  • CNNs, RNNs, GNNs
  • Computer vision
  • Time-series modeling
  • PyTorch
  • TensorFlow
  • scikit-learn
  • MLflow

Generative AI and Agentic Systems

LLM and multi-agent systems grounded in real data, with guardrails and evaluation built in.

  • Vector, graph, and hybrid RAG
  • Multi-agent orchestration
  • MCP
  • Fine-tuning (LoRA, QLoRA)
  • Prompt and context engineering
  • Guardrails
  • Evaluation
  • AWS Bedrock

Data Engineering

Reliable pipelines and data stores for messy, multimodal data.

  • ETL pipelines
  • Apache Spark
  • Kafka
  • PostgreSQL
  • Vector databases
  • Neo4j
  • Databricks
  • Snowflake
  • dbt
  • AWS
  • GCP
  • Azure

Software Engineering and DevOps

From prototype to production, across front end, back end, and cloud.

  • FastAPI
  • Django
  • Flask
  • JavaScript
  • Docker
  • Kubernetes
  • CI/CD
  • Git
  • Agile
  • AI-assisted development

Solution Architecture and Technical Leadership

Setting the technical direction, then building the people, standards, and governance to carry it.

  • Architecture decision records
  • KPI ownership
  • AI strategy advising
  • AI governance
  • Vendor evaluation
  • Executive briefings
  • Code review

Education Technology and Instructional Design

Connecting learning design, data, and AI on a platform serving 65M+ users.

  • Instructional design
  • Learning analytics
  • Educator training
  • Curriculum alignment
  • Content recommendation
  • Content quality models
  • Accessibility
  • AI-assisted content creation

Teaching and Mentoring

Designing and teaching courses at Harvard and the Indian Institute of Science, and mentoring 40+ students.

  • Student mentoring
  • Course design
  • Backward design
  • Assessment design
  • Research-based learning
  • Digital content development
  • Open educational resources

Communication and Public Outreach

Making technical work clear to scientists, leaders, policymakers, and the public.

  • Technical writing
  • Public outreach
  • Stakeholder presentations
  • Fluency in six languages

Media Production

Telling stories through film, design, and AI-generated video.

  • Documentary film
  • Screenwriting
  • Music production
  • Adobe suite
  • AI video generation
  • Text-to-speech
  • Instructional media

Equity and Access in Education

Keeping learning open through free tutoring, guest lectures, and open courseware.

  • Free tutoring
  • K-12 talks
  • Equitable content design
  • Student fundraising
  • Open courseware

About

My vision

I dream of a just world where all life is treated with dignity and growth works with nature instead of against it. I believe making high-quality, relevant education available to everyone is the most sustainable way to get there.

Learning outcomes are falling for much of the world. Too often, what students are taught doesn’t match what they want to learn, or what they will need. Education should spark curiosity, cross subject boundaries and prepare people for real problems, backed by what research tells us about how people learn.

Technology and AI can replace repetitive tasks, not humans, and augment the work people are passionate about. AI opens a new opportunity for a more equitable world, but seizing it would take a new kind of education, for everyone.

The path so far

  1. 2004–09

    B.S. & M.S. in Biology

    University of Calcutta

    Specialized in Zoology, Ecology and Biostatistics.

  2. 2009–17

    Ph.D. in quantitative behavioral sciences

    Indian Institute of Science

    Statistical modeling of learning, memory, decision-making during navigation and searching. Mentored by Raghavendra Gadagkar.

  3. 2016

    Visiting scientist

    CRCA, Université de Toulouse

    Neural and behavioral basis of cooperation in honeybees. Hosted by Martin Giurfa.

  4. 2018–23

    Postdoctoral fellow

    Center for Brain Scince, Molecular and Cellular Biology, Harvard University

    Developed computer vision and deep leqrning pipelines for investigating behavior in large-scale. Mentored by Venki Murthy and L. Mahadevan.

  5. 2020–22 2025-

    Course developer & project leader

    Harvard FAS

    Computational Behavioral Sciences (LS100).

  6. 2023–25

    Data Engineer & AI Solutions Architect

    LabXchange, Harvard University

    Led data and AI/ML strategy for the platform.

  7. 2025–

    Founder

    CViable and Yogi’s Bliss

    Two AI products, in alpha.

Beyond work

The other side of me

Visual arts

Trained as a painter, with a six-year diploma in visual arts. My award-winning photographs have appeared in BMC Ecology and Sanctuary Asia, and I was photographer and head of documentary production for the Indian Institute of Science’s Archives and Publications Cell.

Music

A multi-instrumentalist on tabla, guitar, and harmonica, drawn to the neuroscience of music. I have a varied taste, including Indian classical, blues, and classic rock.

Contact

Let’s talk

I’m always glad to talk about data and AI roles, responsible AI, learning science, or a project we could build together. I also consult on data and AI/ML for research labs and early-stage startups.

Based in Boston, MA, USA

souvik_mandal@harvard.eduConnect on LinkedIn