+:+===**++;;:^^^:=***===+;`^:;;:::;::::^ arsalan@github
=:===##*+;;;:^^^:;:^`````^`^;;;:::;::::^ --------------------------------
;:=*=#%**=;:`'. . .`^:;;;;;::^: Role......: AI Performance Engineer
:;**=##=*=` ....`^:^^```. '^;;:;:::: Company...: ZAL Analytics (@ZAL-Analytics)
:;=*+===+`';=*##%%@@@@@%%%#=:. `:::;;:: Location..: Utrecht Area, Netherlands
^;==;;++`:%@%%@@@@@@@@@@@%%%%*' ^;;:::; Uptime....: 5+ years shipping AI systems
`+==;++:'*%%%%@@@@@@@@@@@@@%%%:. ';:::::
^+==;++^^#%@@@@@@@@@@%#*===#%%;`'.:::::: Kernel....: Rust (C / C++ / CUDA / HIP)
^=*=:^^^:#*;:::;#@@@+``^::^^:#*:`':::::; Runtime...: Vulkan · TensorRT · ONNX · llama.cpp
^==+^^^^::^;++;;=@@@%*=;:;=*##%=::::::^: Focus.....: Low-bit inference · GPU · Air-gapped AI
^+=;^^^:;=*++'';%%@@@%%=``**%%%*=*=;::^^ Stack.....: LoRA · Ontology · RAG · AI Agents
:+++;;;;+=*#%==%@%%%%%@@%%%%%%%==**;::^^
+==*****==*%%%@@%#%%%@@@@@%%%%#+**+:^^^: Repos.....: 34 public (150+ stars, 6k+ downloads)
+++===++++*%%%@@%=**=#%%@@%%%%=;#*^:^^^^ History...: ex Tether · Nscale · AIM Sport · AMD · NXP
=***#*==+++*%%%=:`'`^^^^:*%%#=:^;::::^^^ Edu.......: MSc Data Science · BSc Tech Informatics
****==+=++;^+#;^;=**#***;^*+:^`:::::^^^^ Langs.....: EN (pro) · NL (native) · FR · DE
***=+==*=++;'``^*#+^:+*=:'`''^=*%;:^^^^^ Editor....: nvim · vscode · zed
+===+==*=++=;`..'^`'````...':*##%+^::^::
;++;:;=*=+++**`'....'....':+###*##+::^:: Contact...: arsalan@anwari.nl
:::^^;;=++++*=:::^``''^:;+*###**#=*#==;; LinkedIn..: in/arsalan-anwari
`''`::;::;+=*=;;;;;+++++=#####=++**#%%%#
``'``^:+*#%#*+;;;+;;;++*#**=+++=**####%% Colors....: @@@ ### %%% *** +++ === ::: ...
^:;+*%%%%##*=;::;;;;+;++;;;+==*#########
*#@%%######**+;;:::::;;;+=***######*####
###**####*******=====*******#****#******
AI Engineer with a background in low power and low bit inference, squeezing models onto whatever silicon is in front of me optimized for latency or power usage, from GB200 racks down to low-power embedded boards. Also someone who builds the layer between raw data and AI that reasons over it. Entity resolution, knowledge graphs, RAG and MCP tooling, plus the pipelines, CI/CD and infrastructure to run them in production. Comfortable sitting with domain experts and turning domain knowledge into explicit, testable definitions.
Unlike other AI engineers I care about the parts most people skip: profiling, kernels, portability across heterogeneous back-ends, and making things reproducible.
- Comfortable across the stack: Rust · C · C++ · CUDA · HIP/ROCm · Vulkan · Python
Dutch citizen, 5+ years working across NL, CH, FR, BE and NO.
Languages
GPU / Inference
