Vijay Dachepally has spent more than five years building backend services, distributed systems, and cloud applications. As a Software Development Engineer II at Amazon in Austin, Texas, he works on large scale payment and retail systems used by millions of customers worldwide. His expertise includes Java, Spring Boot, scalable microservices, cloud native architecture, and AWS infrastructure.
That background now powers StylAI, an AI based personal styling platform Vijay developed with Mounika B, an Application Engineer II at Swiss Re. Mounika previously worked at HCA Healthcare, Amazon, and Apps Associates, building experience in backend development, cloud technology, and enterprise applications. Both hold Master of Engineering degrees in Computer Science from the University of Cincinnati, where they met as classmates.
The Problem: Fast Delivery Wasn’t the Real Bottleneck
The idea grew from Vijay’s uncertainty about what to buy and from research the pair conducted into fashion technology in Hyderabad. They encountered a company capable of delivering clothing within 60 minutes, but concluded that faster delivery did not solve the more fundamental problem.
Shopping sites show what is available but rarely understand a shopper’s body, budget, or destination. Pinterest supplies inspiration, usually on someone else, while general AI tools can suggest nonexistent products or unreliable links. Vijay and Mounika believed the missing step was a system that could consider a real person, occasion, region, and budget, then return a complete outfit the user could visualise and purchase.
Dividing the Work: Engineering Meets Fashion Expertise
Development began in May 2026. Vijay led the product vision, architecture, Retrieval Augmented Generation pipeline, AI orchestration, backend infrastructure, analytics, and deployment. Mounika led fashion research, catalogue curation, regional knowledge, and quality validation across Western and Indian markets. Their roles reflect the product’s central challenge: combining reliable engineering with fashion advice that is genuinely wearable.
How the User Experience Works
StylAI asks users for optional details including height, body type, skin tone, budget, and market. They upload a photograph, select from 27 Western and Indian occasions, and receive two outfits with product links and virtual try on images. No signup, account, or paywall is required.
The Five-Stage Process Behind Each Recommendation
Behind that short experience is a five stage process. The photograph is compressed in memory. The platform then searches six specialist knowledge bases for guidance on sizing, dress codes, body proportions, garment construction, skin tone, and colour. Anthropic Claude receives the photograph, profile, and retrieved context and produces two outfits. At the same time, Google Shopping searches for real products and prices while Google Gemini generates the try on images. The platform then assembles the results into shareable cards.
The architecture is designed to degrade gracefully. If shopping data is unavailable, users can still receive the outfits and images. If image generation fails, they can still access the recommendations and product links. A retrieval outage reduces the grounding, but it does not prevent a response.
Why Retrieval Matters: Grounding Advice in Real Data
Retrieval is intended to prevent the model from relying only on general fashion knowledge. When a woman enters a height below 160 centimetres, for example, StylAI classifies her as petite and retrieves proportion guidance covering waist placement, hem length, and silhouette. A man above 183 centimetres is classified as tall and receives different guidance. Sizing information comes from catalogue data, which is especially important in India, where domestic and international brands often follow different conventions.
Localising Recommendations Across Seven Markets
Product selection also reflects geography. StylAI supports seven markets, including India, the United States, Europe, and the United Arab Emirates. Each market produces a localised search with appropriate retailers and currencies rather than a converted price. Budgets apply to the complete outfit, not each item.
Solving for Broken Links and Unrealistic Images
The team learned that direct retailer links could break whenever sites changed their structure. ASOS, Nykaa, and Ajio all did so during development. StylAI now uses verified Google Shopping destinations as a fallback. It also grounds its visual generation in information about fabric and garment construction, since an unrealistically flattering image can encourage a poor purchase.
How StylAI Handles User Data and Privacy
Uploaded photographs are processed in memory and are not written to disk, stored in a cloud bucket, or attached to a permanent user record. StylAI retains anonymous information such as the chosen occasion and market, outfits and brands generated, whether retrieval was active, ratings, and optional feedback. It does not keep names, email addresses, photographs, or raw profiles.
Sharing is optional. If a user shares an outfit card, the generated image is stored behind a random identifier and deleted after seven days. The team also acknowledges that recommendations require sending photographs through encrypted connections to Anthropic and Google. StylAI’s own systems retain no copy when processing ends.
Early Results Since Launch
StylAI entered testing on May 19 and launched publicly on LinkedIn on July 20, 2026, without paid advertising. According to figures supplied by the team through August 14, it recorded 272 unique visitors, 704 sessions, 128 completed styling sessions, and 256 generated outfits. Users opened 300 shopping links, submitted 40 ratings with a 70 percent positive rate, and requested 28 alternate colourways.
The sample remains small, but the shopping activity matters. Users opened more than one product link for every outfit generated. Date nights were the most requested occasion, followed by casual wear and parties, suggesting the service may be useful for routine decisions rather than only formal events.
What’s Next for StylAI
The roadmap centres on making StylAI behave more like an ongoing stylist. Planned features include recommending combinations from clothes users already own, saving outfits, remembering preferences across conversations, and expanding to mobile. StylAI remains free, with feedback rather than monetisation as the current priority.
Part of a Broader Research Ambition
The project also supports Vijay’s broader research ambitions. He has reviewed approximately 45 papers, primarily in artificial intelligence and machine learning, and is authoring research on RAG-based personalised styling recommendations. His long term goal is to become a recognised expert in software engineering, distributed systems, cloud computing, and applied AI. StylAI gives Vijay Dachepally a working example of how those disciplines can be combined to make consumer AI more grounded, useful, and personal.
