Keenan Pereira
AI Martech Engineer · Data · ML · Web
Profile
AI Martech Engineer at AHA Smart Homes, Mumbai. I own the website and search programme, the marketing data warehouse and the AI tooling around them. In the ten weeks after I took over in August 2026, non-brand Google clicks rose 183%. BSc (Hons) in AI & ML; earlier roles in Gen AI, ML and data science. I build with AI coding agents and check every number before it ships.
Experience
AHA Smart Homes · AI Martech Engineer
- Run the website and search programme since Aug 2026: technical SEO overhaul, a content system with 43 articles of my own and 12× more structured data. Non-brand Google clicks +183% and average non-brand rank 18.5 → 9.2 in ten weeks.
- Built a read-only marketing data warehouse (Zoho CRM, Google Ads, Meta, call logs) that syncs every 15 minutes. An independent audit fixed a 2× lead double count and spend that had gone unrecorded.
- Shipped a phone-first consultant dashboard showing the gap to target, a call-first list and referral prompts, with start-up time cut 67%.
- Built the showroom TV app and deals map: a pitch that plays itself (about 5½ minutes, was 15), a path-traced 3D tour of the experience centre, and every AHA building and billboard site on one map, tuned for a 2 GB TV.
- Made lead tracking trustworthy: save verification across 7 forms (21/21 Playwright tests), two silently broken ad forms fixed, and 90-day attribution going live.
- Mapped every home-automation firm in Mumbai and across India with the Google Places API: a quadtree sweep past the 60-result cap found 1,176 firms in Mumbai and 11,007 nationally.
Artificial Intelligence Engineer Intern
- Built AHA’s first data platform (nightly CRM sync, sales-call mining, a weekly ads agent, a BI router answering in under 1 s) and an ads landing page with 94% lighter images.
Unreal Dimension Studio · Gen AI Engineer
- Engineered and deployed production-ready ML models that improved product functionality and performance.
- Built agentic AI systems with LangChain and local models through Ollama.
- Streamlined the end-to-end ML pipeline, from research and model development to integration and deployment.
EdTechSociety · Machine Learning Engineer Intern
- Fine-tuned a CNN to detect pneumonia from chest X-rays, using class weighting for a 1:3 imbalance. 99.2% recall, 74.7% test accuracy.
- Built a live emotion-detection web app: a Keras CNN trained on FER-2013 (7 classes), served with Flask from the webcam.
NeoSOFT · Data Science Intern
- Credit-score classification on 100k monthly records: per-customer cleaning, RFE and four models compared. Random Forest 76.6%, macro F1 0.75.
- Forecasting with LSTM and Prophet (NIFTY 50) and SARIMAX (store demand), and a Gemini app that answers questions about an uploaded CSV.
Digirise Infolabs · Data Science Intern
- Led projects from analysis to model: classifiers, sales regression, credit-card fraud detection and stock time series.