PRISM LABSResearch studying the impact of major policy announcements on financial-market behavior and volatility.
- 8th Pan-IIT International Conference on Financial Markets & Corporate Finance
- 03—05 Jul 2026
PRISM LABS
Building at the intersection of intelligence, technology and real-world problems.
AI research lab · Est. 2026
Research informs what we build. Building gives research a real-world test.
Prism Labs is an independent AI research and engineering lab, founded in June 2026. We explore interdisciplinary research, build deep-tech systems, and test ideas against meaningful real-world problems.
Our work is not limited to papers or prototypes. Where an idea proves useful, we aim to develop a working system, validate it in context, and pursue a commercial application only when it offers genuine value.
01 / The problem
Advertising is still largely separated from the content people consume.
TV commercials and social media advertisements compete for attention outside the content itself. Films, however, already contain natural advertising surfaces — objects that exist inside the visual world of the story. The challenge is recognizing those surfaces reliably and adapting them without breaking the scene, story, or viewing experience.
- A billboard.
- A bus.
- A storefront.
- A digital screen.
- A poster.
Why not make those surfaces dynamic?
02 / What we’re building
The same movie. The same scene. Different advertisement.
Prism Labs is researching computer-vision and media-infrastructure techniques that could identify contextual advertising opportunities in films and other visual media, then adapt an in-scene surface to factors such as geography, audience, campaign, and placement. The same film could potentially carry different advertising content depending on where and how it is shown — while the story and scene remain unchanged.
One film↓
One scene↓
One advertising surface↓
Multiple contextual campaigns
This is an early-stage research and prototyping direction. Potential users could include film studios, streaming or distribution platforms, and advertisers, but Prism Labs has no deployed product, customers, or commercial commitments in this area today.
Research → Build → Validate → Deploy
01
Research
Understand the problem.
02
Build
Turn ideas into working systems.
03
Validate
Test assumptions against evidence.
04
Deploy
Move useful technology into the real world.
03 / Research
01 / 03
Presented research on cybersecurity and privacy challenges in federated learning for medical imaging.
- CAICS 2026
- IISc Bengaluru
Funded by the U.S. Department of State. Jointly hosted by Argonne National Laboratory and IISc Bengaluru.
View LinkedIn post ↗Research on selective state space models and their potential as a linear-time alternative to attention.
- Accepted · NeurIPS 2026 Education Track
- Sydney, Australia
Computer-vision research applying an scSE-augmented SegFormer to segmentation in historical Telugu palm-leaf manuscripts.
- Accepted · CVIP 2026
Research becomes useful when it survives contact with reality.
01Observe
02Experiment
03Validate
04Build
04 / People
Deepak Kambala
Founder — Prism Labs, DeepTech ResearchFounder, Geeks Hive · Computer Science undergraduate
Selected evidence
- Accepted to NeurIPS 2026 Education Track for research on selective state space models.
- Research presented at CAICS 2026 at IISc Bengaluru and ICFMCF 2026.
- Global Rank 28 in the ICML 2026 Agent Reproductions Challenge.
- Winner, Square Hacks 2025 at IIT Delhi; runner-up, SysML 2.0 Hackathon at IISc Bengaluru.
Program credentials
Prism Labs has been accepted into startup programs by Sarvam, MongoDB, and GitLab. These are program credentials, not claims of funding, partnership, customer traction, or product deployment.
View founder portfolio ↗Have a problem worth investigating?
We are interested in thoughtful conversations about research, engineering, and difficult real-world problems. Prism Labs is open to exploring other AI and deep-tech directions as its capabilities evolve.
Start a conversation ↗