Machine-learning engineer working across production agents, retrieval, privacy-aware ML, MLOps, geospatial planning, forecasting and multimodal research.
Lead architecture and principal engineering across multi-agent commerce workflows, telecom recommendations, retrieval evaluation, privacy controls and cloud data systems. Disclosure-safe examples include a production order-and-fulfilment agent adapted across two US retail brands and a two-layer PII detection pipeline for a Canadian property group.
FastAPI, Next.js and deck.gl planning environment integrating five data and modelling layers. Presented at the Asian Development Bank Asia Clean Energy Forum 2025.
ESP8266 to AWS IoT Core, DynamoDB and Lambda ingestion with AQI processing and IDW heatmaps. Diagnosed a months-long silent stream failure and rebuilt the observability path. Programme spanning eight university partners.
Agentic RAG across a 150-page policy roadmap with section-level citations and unsupported-claim controls.
Urdu and English PWA for fuel and EV cost comparison, deployed through kiosk and QR interactions at a national expo.
Battery-sizing, efficiency and thermal models for a convertible motorcycle plug-in hybrid. Public campaign reached approximately seven million views.
Long-horizon memory for LLM agents. Naive RAG, Mem0, A-MEM and MEM1 evaluated on LongMemEval and LoCoMo; reinforcement-learning controller for short- and long-term memory management.
Contamination-resistant behavioural benchmark for extended agent interaction.
LLM and RAG interaction layer, ADL and gait classification, rPPG vital estimation and multimodal agitation/risk detection. Co-author of a 2024 JMIR Research Protocols study protocol.
RAG pipelines and graph analytics over open-source reporting as a nonresident research associate at a university policy research centre.
Research focus: long-horizon memory and behavioural evaluation for LLM agents.
Signals, control systems, embedded engineering and autonomous systems.
AI for energy distribution and EV infrastructure.
Renewable energy, EV charging and BESS optimisation.
Co-author of the protocol for an empathetic patient-robot interaction study addressing agitation and fall risk among people with dementia.
Agent orchestration · retrieval and evaluation · PyTorch · FastAPI · forecasting · geospatial analytics · computer vision · reinforcement learning · AWS · Azure · Docker · data orchestration · edge inference