TechnoLynx
Budapest, Hungary
About
TechnoLynx is a Budapest AI engineering firm working on computer vision, GPU inference optimization and generative systems for regulated and hardware-constrained deployments.
Who They Work With
Hardware, codec and vision startups that need a model or GPU pipeline made faster, ported or validated, with a benchmark they keep.
Based on 4 client engagements with a source URL · researched 23 Sept 2026
Our Take
The retail vision studies publish their architecture and their test sets: the autonomous-cart SKU recognizer names a frozen DINO encoder, data-derived superclasses and a few-shot head, reporting above 95% accuracy at about 1,000 classes and above 83% at about 2,000, and the barcode pipeline beat a Dynamsoft baseline 86.7% to 80% on the same 30-video set.
The one named AI engagement that reached a launched product is Tasty Tech's Dentrino On-Demand, a second-generation dental image-synthesis product; the cart client stays anonymous with its barcode edge hardware not finalized, and the fraud and anomaly-detection studies end in an audit and a handed-over proof of concept.
The V-Nova and CloudRF studies, two of the named-client pieces, are GPU performance engineering with no model in them—credible work, but not AI delivery.
The weaker fit is buyers who want a vendor to run a model in production after handover: the packs are built to end in a benchmark, eval harness or runbook your team keeps, and the team page names four engineers alongside the founder.
Key strengths
- Studies publish test conditions next to results—a 30-video test set, a Dynamsoft baseline on identical footage, exact top-1 accuracy with top-k removed—so a number can be checked, not only quoted
- Founder Balázs Keszthelyi is a named contributor to the SYCL specification and ran advanced R&D at V-Nova, whose CEO is quoted on the homepage as a client
- The write-ups admit what did not work: the fraud audit says generated samples lacked the fraud properties the client cared about, and the dental study says TensorFlow might have suited delivery better
Good to know
- The deepest vision work on the site names no client—ask to speak to the autonomous-cart customer and whether the multi-year perception system runs in stores today
- Several engagements end in a proof of concept or audit the client took over—if you need a vendor to operate the model, confirm that scope before signing a pack
- The site takes one client per technology niche—ask early whether a competitor already holds yours, before investing time in scoping
Reviewed by Gabor Kiss
Founder & Curator, AIAgencies.eu · 23 Sept 2026
Based on the agency's publicly available portfolio. Own this agency? Claim your profile to provide additional context or request a review update.
Notable Clients
Every engagement documented in a published case study
Documented engagements: Tasty Tech (Dentrino), V-Nova, CloudRF, and Kineon. All 4 carry a source URL on file.
Services
Industries
Get Matched with Agencies
RecommendedTell us your needs and we'll connect you with 3 vetted agencies—including TechnoLynx if it's a fit.
Get Matched →Curator shortlist within 3 business days
Visit TechnoLynx's website →Own this agency? Claim your profile to verify info & receive leads directly.
