The product problem
Language practice needs continuity, comprehensible support and opportunities to respond. A translated paragraph alone does not provide the same experience as aligned text, pronunciation, speech and context.
Engineering & product decisions
Keep assistance in the conversation
Practice-language replies sit alongside native-language layers, replay and review so people can keep participating while seeking help.
Treat alignment as an interface contract
Transliteration normalization checks source coverage, grapheme boundaries and script-specific units before displaying aligned support.
Preserve visual context across devices
Conversational camera work handles capture identity, durable media, replay, deletion and native/browser lifecycle differences.
Build organizational operations alongside the experience
Implemented organization administration, usage reporting and billing support deployment discussions. They are capabilities, not evidence of enterprise adoption.
What the evidence supports
The public web route and reviewed demo are live. Repository evidence supports implemented capabilities; no customer adoption, learning outcome or enterprise certification is claimed. Generated language still requires quality review. Source is private.
This case study describes Arjun Iyer’s product and engineering work through Fallible Software, Co. Model services, frameworks and upstream runtimes retain their own provenance. Evidence and status notes →
Discuss this kind of work ↗