Deep
Learning
Technical
Skills
01
Mathematical
Foundations
02
Neural
Network
Fundamentals
03
Model
Architectures
04
NLP
&
Large
Language
Models
05
Computer
Vision
&
Multimodal
06
Data
&
Feature
Engineering
07
Training
at
Scale
08
MLOps
&
Deployment
09
Evaluation
,
Safety
&
Governance
01
Mathematical
Foundations
12
S KILLS
Linear
algebra
·
vectors
,
matrices
,
tensor
decomposition
Multivariable
calculus
·
Jacobians
,
Hessians
,
chain
rule
Probability
&
statistics
·
Bayes
,
MLE
,
MAP
Optimization
theory
·
convex
and
non
-
convex
Gradient
descent
variants
·
SGD
,
momentum
,
AdamW
Information
theory
·
entropy
,
KL
divergence
Automatic
differentiation
·
forward
and
reverse
mode
Eigen
-
decomposition
&
PCA
Dimensionality
reduction
·
t
-
SNE
,
UMAP
Numerical
stability
·
precision
,
conditioning
Sampling
methods
·
Monte
Carlo
,
MCMC
Bayesian
inference
·
variational
inference
02
Neural
Network
Fundamentals
12
S KILLS
Multilayer
perceptrons
&
feedforward
networks
Backpropagation
&
computational
graphs
Activation
functions
·
ReLU
,
GELU
,
SiLU
Loss
functions
·
cross
-
entropy
,
contrastive
,
triplet
Weight
initialization
·
Xavier
,
He
,
orthogonal
Regularization
·
dropout
,
weight
decay
,
label
smoothing
Normalization
·
batch
,
layer
,
group
,
RMSNorm
Residual
&
skip
connections
Learning
-
rate
schedules
·
warmup
,
cosine
,
one
-
cycle
Hyperparameter
tuning
·
Optuna
,
Bayesian
search
Gradient
clipping
&
exploding
gradients
Early
stopping
&
model
checkpointing
1 / 3
03
Model
Architectures
12
S KILLS
Convolutional
networks
·
ResNet
,
VGG
,
EfficientNet
Vision
Transformers
·
ViT
,
Swin
,
DINOv
2
Recurrent
networks
·
LSTM
,
GRU
,
bidirectional
RNN
Transformers
·
self
-
attention
,
multi
-
head
attention
Encoder
–
decoder
&
sequence
-
to
-
sequence
Autoencoders
&
variational
autoencoders
Generative
adversarial
networks
·
StyleGAN
,
CycleGAN
Diffusion
models
·
DDPM
,
latent
diffusion
,
flow
matching
Graph
neural
networks
·
GCN
,
GAT
,
message
passing
State
space
models
·
S
4,
Mamba
Mixture
-
of
-
Experts
routing
Neural
architecture
search
04
NLP
&
Large
Language
Models
12
S KILLS
Tokenization
·
BPE
,
WordPiece
,
SentencePiece
Embeddings
·
Word
2
Vec
,
GloVe
,
sentence
transformers
Transfer
learning
&
fine
-
tuning
Parameter
-
efficient
tuning
·
LoRA
,
QLoRA
,
adapters
Instruction
tuning
&
RLHF
Preference
optimization
·
DPO
,
PPO
,
KTO
Prompt
engineering
&
in
-
context
learning
Retrieval
-
augmented
generation
(
RAG
)
Vector
databases
&
semantic
search
Long
-
context
attention
·
RoPE
,
ALiBi
,
sliding
window
Inference
quantization
·
GPTQ
,
AWQ
, 4-
bit
Decoding
strategies
·
beam
,
top
-
k
,
nucleus
05
Computer
Vision
&
Multimodal
10
S KILLS
Image
classification
·
ImageNet
-
scale
training
Object
detection
·
YOLO
,
Faster
R
-
CNN
,
DETR
Segmentation
·
U
-
Net
,
Mask
R
-
CNN
,
SAM
Image
generation
·
Stable
Diffusion
,
ControlNet
Video
understanding
&
temporal
modeling
Pose
estimation
&
keypoint
detection
OCR
&
document
understanding
Vision
–
language
models
·
CLIP
,
BLIP
-2,
LLaVA
Text
-
to
-
image
&
text
-
to
-
video
synthesis
3
D
vision
·
NeRF
,
Gaussian
splatting
,
point
clouds
06
Data
&
Feature
Engineering
10
S KILLS
Dataset
curation
&
cleaning
Data
augmentation
·
mixup
,
CutMix
,
RandAugment
Synthetic
data
generation
Labeling
workflows
&
annotation
QA
Class
imbalance
·
focal
loss
,
resampling
Feature
engineering
&
embedding
pipelines
Data
versioning
·
DVC
,
lakeFS
Batch
&
streaming
pipelines
·
Spark
,
Kafka
Split
design
&
leakage
prevention
Federated
learning
&
differential
privacy
2 / 3
07
Training
at
Scale
10
S KILLS
Distributed
training
·
DDP
,
FSDP
,
DeepSpeed
Parallelism
·
tensor
,
pipeline
,
expert
Mixed
precision
·
FP
16,
BF
16,
FP
8
Gradient
accumulation
&
micro
-
batching
CUDA
programming
&
custom
kernels
GPU
/
TPU
cluster
orchestration
Communication
optimization
·
NCCL
,
compression
Memory
optimization
·
ZeRO
,
activation
checkpointing
Profiling
·
PyTorch
Profiler
,
Nsight
Experiment
tracking
·
W
&
B
,
MLflow
,
TensorBoard
08
MLOps
&
Deployment
12
S KILLS
Model
serving
·
TorchServe
,
Triton
,
vLLM
Containerization
·
Docker
,
Kubernetes
CI
/
CD
for
ML
·
GitHub
Actions
,
Kubeflow
Model
registry
&
versioning
Feature
stores
·
Feast
,
Tecton
Monitoring
&
drift
detection
A
/
B
testing
&
shadow
deployment
Edge
inference
·
TensorRT
,
ONNX
Runtime
,
TFLite
Model
compression
·
pruning
,
distillation
,
quantization
Latency
optimization
·
batching
,
KV
caching
Autoscaling
&
cost
optimization
Reproducibility
&
infrastructure
as
code
09
Evaluation
,
Safety
&
Governance
10
S KILLS
Metrics
·
precision
,
recall
,
F
1,
mAP
,
BLEU
,
perplexity
Benchmarks
·
MMLU
,
GLUE
,
COCO
,
ImageNet
Error
analysis
&
ablation
studies
Interpretability
·
SHAP
,
Grad
-
CAM
,
attention
maps
Fairness
&
bias
auditing
Robustness
&
adversarial
testing
Hallucination
detection
&
factuality
checks
Safety
alignment
&
red
-
teaming
Model
cards
&
documentation
Governance
·
EU
AI
Act
,
NIST
AI
RMF
Deep
Learning
Technical
Skills
· 100
skills
across
9
categories
.
Compiled
as
a
reference
sheet
for
résumé
writing
,
hiring
scorecards
,
curriculum
planning
,
and
self
-
assessment
.
3 / 3
