Founder of Wisko · AI Researcher & Engineer
I work at the intersection of large language models, computer vision, and real-time voice AI. My undergraduate research built a Vision Transformer for plant-disease detection, and I now lead Wisko, a multi-tenant agentic SaaS platform. I'm seeking graduate research opportunities to deepen this work.
My company. Wisko is a multi-tenant SaaS platform where autonomous agents generate complete whiteboard-animation videos: applied agentic AI and multimodal generation running in production.
01 · About
I'm an AI engineer with a physics background and a strong pull toward research. My undergraduate final-year project designed an Object Detection Transformer and an AI digital microscope to detect cotton-crop diseases from visual symptoms, bridging computer vision and real-world agricultural sensing.
Professionally, I build production LLM and voice-AI systems: multilingual real-time translation across 30+ languages, agentic workflows, and efficient inference via fine-tuning, KV caching, and quantization. I'm eager to bring this systems-and-research experience into a graduate program focused on LLMs, multimodal learning, or computer vision.
02 · Education
CGPA: 3.63 / 4.00 (equivalent to a GPA of 3.63 on the US 4.0 scale) · EQF Level 6
Thesis: “Object Detection Transformer for disease detection on cotton leaves.” Field: Computational Physics, computer programming, mathematics & statistics.
Languages
03 · Research & Projects
Final Year Research Project · Computer Vision
Designed an Object Detection Transformer and an AI-modeled digital microscope to detect cotton-crop diseases from visual symptoms, outperforming traditional CNN baselines in classification accuracy.
Freelance research study · UK
A benchmarking study carried out for a UK-based student that evaluated a wide range of computer-vision models for detecting vehicles at airports. The core question was whether camera and vision-based systems could replace LiDAR by estimating depth, distance, and detection quality across changing lighting, seasons, and weather.
GovTech Research · Speech & LLMs
Bi-directional voice translation across 30+ languages at sub-1500 ms latency, combining Whisper ASR, custom speech pipelines, and context-aware memory for cross-language consistency.
Retrieval-Augmented Generation
A privacy-preserving document assistant on local Llama 3 and Qdrant, enabling semantic search over large corpora with no external data leakage.
Agentic Systems · CrewAI
A swarm of autonomous agents that gather information, run sentiment analysis, and synthesize structured reports, featuring self-correction loops and tool use.
Embedded Computer Vision
Deployed lightweight YOLOv8 models on Jetson Orin Nano for real-time detection at <40 ms fully-offline inference in low-bandwidth environments.
Instrumentation & Sensing
Portable Raspberry Pi radar for remote sensing with real-time signal processing, plus IoT safety systems including a blind-spot accident-prevention module and gas-leakage detector.
Multimodal ML · VLM
Fine-tuned Llama 3.2 Vision with Unsloth on radiography imagery, adapting a vision-language model to medical image understanding.
LLMOps · Unsloth + LoRA/QLoRA
An efficient fine-tuning pipeline using Unsloth with LoRA and QLoRA for domain-specific LLM adaptation on constrained hardware.
Diffusion models · Inference
An open-source pipeline that exports Stability AI's Stable Audio 3 Small (music & SFX) diffusion transformer to ONNX, so it runs CPU/GPU inference without a PyTorch runtime dependency.
Agentic AI · LLM orchestration
An AI assistant built with LLM orchestration, tool use, and conversational memory for multi-step, context-aware reasoning.
Multimodal systems · LangGraph
A LangGraph-orchestrated pipeline generating script → TTS → images → whiteboard sketch animation, stitched with FFmpeg and served via FastAPI + APScheduler on Supabase & GCS, running as a full multimodal generation system.
Code & write-ups on GitHub →
04 · Skills
05 · Experience
06 · Certifications
07 · Service & Leadership
ICPSE-2021 & ICPSE-2022 (International Conferences on Physical Sciences & Engineering) and ICSMAND (9th International Conference on Semiconductor Materials & Nano-Devices).
President, Natural Resources Defense Council & Society, KFUEIT.
Organized the 1st South Punjab Science Mela 2021; participant in the World Water Day webinar on renewable-energy water electrolysis.
Student Proctor at the 1st KFUEIT Convocation; organized a Blood Donation Camp at Sehja Public Health Care Centre.
08 · Contact
I'm seeking graduate research and MS/PhD opportunities in AI. I'd be glad to share my full CV, transcripts, or code. The best way to reach me is email. Copy it below, or click to open your mail app.
osamaaltaf.pk@gmail.com