Mike369williams commited on
Commit
908b171
·
verified ·
1 Parent(s): e9af4d9

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +131 -0
README.md CHANGED
@@ -81,6 +81,137 @@ To build India’s most practical, multilingual AI model optimized for:
81
 
82
  ---
83
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
84
  ## 📩 Contact
85
  Founder: **Srikanth**
86
  Email: **boorgalasrikanth@gmail.com**
 
81
 
82
  ---
83
 
84
+ ## 📈 Market Opportunity
85
+
86
+ India has 1.4 billion users across dozens of languages, yet most AI models are optimized for Western datasets.
87
+ Sanchari focuses on:
88
+
89
+ Indian English, Telugu, Hindi
90
+
91
+ Local accents
92
+
93
+ Local knowledge
94
+
95
+ Culturally aligned reasoning
96
+
97
+ Vernacular business workflows
98
+
99
+
100
+ Target Markets:
101
+
102
+ Enterprises adopting AI
103
+
104
+ Customer support automation
105
+
106
+ Healthcare conversational assistants
107
+
108
+ FinTech support & KYC automation
109
+
110
+ Education & e-learning
111
+
112
+ Government services (Digital India)
113
+
114
+
115
+ Projected TAM (India AI Assistants): $3.5B+ by 2027
116
+
117
+
118
+ ---
119
+
120
+ ## ⚡ Competitive Advantage
121
+
122
+ Sanchari is designed specifically for Indian users, unlike global models trained mostly on Western data.
123
+
124
+ Key differentiators:
125
+
126
+ Native support for Telugu + Hindi + Indian English
127
+
128
+ Dataset curated for Indian knowledge, culture, and business workflows
129
+
130
+ Lightweight model versions for on-device and low-compute deployment
131
+
132
+ Faster inference
133
+
134
+ Lower cost for Indian startups
135
+
136
+ Can be embedded into apps & enterprise workflows
137
+
138
+ Privacy-friendly deployment options
139
+
140
+
141
+
142
+ ---
143
+
144
+ ## 🔧 Technical Architecture (High-Level)
145
+
146
+ Tokenizer
147
+
148
+ Multilingual tokenizer optimized for Indic languages
149
+
150
+ Handles mixed-script text (Eng + Indic)
151
+
152
+
153
+ Model Family
154
+
155
+ Sanchari-S (200–350M) — prototype
156
+
157
+ Sanchari-M (1–3B) — mid-range
158
+
159
+ Sanchari-L (7B+) — flagship foundation model
160
+
161
+
162
+ Training Stack
163
+
164
+ PyTorch + DeepSpeed
165
+
166
+ FlashAttention
167
+
168
+ LoRA adapters for efficient instruction tuning
169
+
170
+ Multi-GPU distributed training
171
+
172
+
173
+
174
+ ---
175
+
176
+ ## 💰 Funding Plan (Seed: ₹25,00,000)
177
+
178
+ Where the funds go:
179
+
180
+ Category Cost
181
+
182
+ Multilingual licensed datasets ₹6,00,000
183
+ Compute for training S, M models ₹12,00,000
184
+ Storage, inference, and deployment ₹3,00,000
185
+ Evaluation, safety testing ₹1,00,000
186
+ Team & operations ₹3,00,000
187
+
188
+
189
+ Deliverables to Investors:
190
+
191
+ Checkpoints for Sanchari-S and M
192
+
193
+ Evaluation results
194
+
195
+ Demo API
196
+
197
+ Weekly updates
198
+
199
+
200
+
201
+ ---
202
+
203
+ ## 👤 Founder
204
+
205
+ Srikanth B.
206
+ AI & product innovator focused on practical, multilingual AI solutions for India.
207
+ Experience across product development, engineering leadership, and AI adoption for scalable business use cases.
208
+
209
+ Email: boorgalasrikanth@gmail.com
210
+
211
+
212
+ ---
213
+
214
+
215
  ## 📩 Contact
216
  Founder: **Srikanth**
217
  Email: **boorgalasrikanth@gmail.com**