ai engineer · miami, florida
Hi, I'm Mohith Kanala.
the work
Grounded answers, in 20+ languages, in under 5 seconds.
Built the retrieval backbone for customer and vendor AI assistants: semantic search, custom cross-encoder reranking, regional filters, caching, live inventory context, and production guardrails.
Python · Go · AWS Bedrock · Weaviate · Elasticsearch · NeMo Guardrails
~1 TB managed · 5-10 GB daily · NDCG + Recall@K
Forecasts across 15,000-20,000 locations.
Architected a near real-time forecasting platform connecting ERP, SCM, POS, ordering, and vendor systems. An XGBoost + LSTM ensemble turned 100M+ historical records into location-level decisions.
XGBoost · LSTM · SageMaker · Kafka · Lambda · SQL
8-12% more accurate · 6-8% fewer stockouts · 100-300 GB daily
Autonomous BI operations, with approval built in.
Created an autonomous operations layer spanning 30+ servers and 100+ workflows. It detects anomalies, simulates remediation in a sandbox, and routes structured recommendations for human approval.
Vertex AI · GCP · Power BI · PostgreSQL · Docker · Guardrails AI
60% faster resolution · 40% less manual work · 50-80 signals daily
Automation that learns every 6 to 12 hours.
Built an end-to-end Google Ads system combining live performance, trend signals, LLM-generated creative, audience targeting, and an approval-aware optimization loop.
Python · Google Ads API · Vertex AI · Cloud Functions · Docker
15-30% higher CTR · 50-65% less setup · 20+ campaigns